Makes fasterrcnn backbone run

[NoraHagmeyer]

• Adds tensor.insert_slice() support to the mlir frontend
• Adds a fasterrcnn_resnet50 benchmark to the repository. Only the backbone can be lowered to torch-mlir. Therefore, no harness and benchmarking infrastructure is used for this benchmark.

[daisytuner]

Daisytuner Report - mlir_torch_models (chamomile)

@@                                   Benchmarks                                   @@
=====================================================================================
  Benchmark              Time        ΔTime       Thr         Energy      ΔEnergy     
=====================================================================================
# bn_conv_bn_relu_maxpool_torch18.62 s     +0.51%      N/A         3487.65 J   -3.87%      
# bn_conv_bn_relu_maxpool_run_none3.26 s      +2.77%      N/A         632.98 J    -0.97%      
# bn_conv_bn_relu_maxpool_run_sequential3.29 s      +3.37%      N/A         642.20 J    -0.02%      
# bn_conv_bn_relu_maxpool_run_openmp3.37 s      +3.81%      N/A         660.00 J    -0.89%      
# bn_conv_bn_relu_maxpool_run_cuda3.71 s      +1.05%      N/A         697.14 J    -4.00%      

[daisytuner]

Daisytuner Report - mlir_torch_layers (chamomile)

@@                                   Benchmarks                                   @@
=====================================================================================
  Benchmark              Time        ΔTime       Thr         Energy      ΔEnergy     
=====================================================================================
# batchnorm_torch        19.01 s     -1.23%      N/A         3740.63 J   +0.32%      
# batchnorm_run_none     3.87 s      -0.17%      N/A         761.27 J    +0.58%      
# batchnorm_run_sequential3.98 s      +0.35%      N/A         784.66 J    +1.49%      
# batchnorm_run_openmp   3.51 s      -1.55%      N/A         729.36 J    -0.13%      
# batchnorm_run_cuda     5.47 s      -1.94%      N/A         1078.82 J   -0.47%      
# conv2d_torch           18.59 s     +0.11%      N/A         3652.26 J   +1.16%      
# conv2d_run_openmp      4.34 s      +0.04%      N/A         1039.37 J   +1.88%      
# conv2d_run_cuda        7.54 s      +1.82%      N/A         1474.85 J   +3.11%      
# linear_torch           6.20 s      -0.85%      N/A         1489.29 J   +0.67%      
# linear_run_none        10.52 s     -0.09%      N/A         2890.40 J   +0.38%      
# linear_run_sequential  8.89 s      -0.75%      N/A         2569.07 J   +0.10%      
# linear_run_openmp      8.46 s      +1.81%      N/A         2482.34 J   +2.67%      
# linear_run_cuda        8.38 s      -0.25%      N/A         1635.44 J   +0.58%      
# matmul_torch           6.08 s      +0.21%      N/A         1463.57 J   +1.05%      
# matmul_run_none        10.65 s     +1.25%      N/A         2921.20 J   +1.33%      
# matmul_run_sequential  8.89 s      +0.69%      N/A         2578.28 J   +1.56%      
# matmul_run_openmp      8.28 s      -0.58%      N/A         2432.14 J   -0.02%      
# matmul_run_cuda        8.23 s      +0.32%      N/A         1615.04 J   +1.59%      
# pooling_torch          26.16 s     +1.48%      N/A         5225.36 J   +2.81%      
# pooling_run_none       15.27 s     -0.73%      N/A         2946.75 J   +0.53%      
# pooling_run_sequential 15.19 s     -1.28%      N/A         2935.33 J   +0.11%      
# pooling_run_openmp     8.99 s      -1.87%      N/A         1833.85 J   -0.80%      
# pooling_run_cuda       20.03 s     -0.83%      N/A         3908.24 J   +0.46%      
# relu_torch             19.26 s     +1.68%      N/A         3784.56 J   +2.88%      
# relu_run_none          3.80 s      +0.01%      N/A         750.72 J    +1.30%      
# relu_run_sequential    3.81 s      -0.28%      N/A         751.30 J    +0.77%      
# relu_run_openmp        3.53 s      +1.94%      N/A         725.23 J    +3.28%      
# relu_run_cuda          5.48 s      -0.79%      N/A         1081.98 J   +0.60%      

[daisytuner]

Daisytuner Report - python_npbench (zinnia)

@@                                   Benchmarks                                   @@
=====================================================================================
  Benchmark              Time        ΔTime       Thr         Energy      ΔEnergy     
=====================================================================================
# adi_numpy              1.31 s      -0.43%      N/A         130.82 J    -0.51%      
- adi_omp                14.61 s     +82.84%     N/A         1422.99 J   +79.08%     
- adi_cuda               4.69 s      +10.03%     N/A         454.05 J    +9.55%      
- adi_seq_tuning         15.13 s     +83.89%     N/A         1401.41 J   +83.29%     
# atax_numpy             2.15 s      -0.08%      N/A         222.84 J    -0.14%      
# atax_omp               2.96 s      -1.22%      N/A         370.55 J    -1.82%      
# atax_cuda              4.14 s      +0.74%      N/A         424.87 J    +0.63%      
# atax_seq_tuning        4.10 s      -0.44%      N/A         397.79 J    +0.08%      
# gemm_numpy             1.22 s      +0.95%      N/A         195.16 J    +0.80%      
# gemm_omp               1.12 s      +0.56%      N/A         162.94 J    +0.15%      
# gemm_cuda              10.63 s     +0.40%      N/A         1011.22 J   +0.48%      
# gemm_seq_tuning        1.11 s      -0.10%      N/A         161.69 J    -0.28%      
# gesummv_numpy          1.75 s      -0.49%      N/A         249.64 J    -0.51%      
# gesummv_omp            1.96 s      -1.39%      N/A         308.00 J    -1.87%      
# gesummv_cuda           8.34 s      +0.13%      N/A         1001.69 J   +0.08%      
# gesummv_seq_tuning     8.58 s      -0.51%      N/A         974.64 J    -0.03%      
# gemver_numpy           1.10 s      +0.98%      N/A         169.15 J    +1.11%      
# gemver_omp             843.10 ms   +0.02%      N/A         107.36 J    -0.37%      
# gemver_cuda            3.90 s      +0.95%      N/A         390.24 J    +0.91%      
# gemver_seq_tuning      5.51 s      -0.49%      N/A         496.62 J    +0.59%      
# k2mm_numpy             1.20 s      -0.15%      N/A         196.40 J    -0.11%      
# k2mm_omp               3.57 s      -0.43%      N/A         662.22 J    -0.62%      
# k2mm_cuda              13.59 s     -0.13%      N/A         1290.27 J   +0.17%      
# k2mm_seq_tuning        2.91 s      -3.09%      N/A         390.51 J    -1.50%      
# k3mm_numpy             1.03 s      -0.20%      N/A         181.82 J    -0.24%      
# k3mm_omp               5.55 s      +0.03%      N/A         947.43 J    -0.85%      
# k3mm_cuda              19.79 s     -0.07%      N/A         1864.91 J   -0.14%      
# k3mm_seq_tuning        4.90 s      -0.90%      N/A         686.17 J    -0.17%      
# mvt_numpy              2.43 s      -0.09%      N/A         247.61 J    -0.17%      
# mvt_omp                2.74 s      -0.23%      N/A         284.54 J    -0.19%      
# mvt_cuda               3.36 s      -0.32%      N/A         342.50 J    -0.23%      
# mvt_seq_tuning         2.74 s      +0.01%      N/A         284.26 J    +0.02%      
# symm_numpy             794.75 ms   -0.57%      N/A         81.74 J     -0.44%      
- symm_omp               6.06 s      +29.35%     N/A         595.37 J    +28.66%     
- symm_seq_tuning        8.31 s      +20.20%     N/A         749.91 J    +20.80%     
# syr2k_numpy            894.33 ms   -0.11%      N/A         90.93 J     -0.02%      
- syr2k_omp              9.85 s      +35.42%     N/A         935.82 J    +34.66%     
# syr2k_cuda             1.64 s      +7.62%      N/A         170.26 J    +6.89%      
- syr2k_seq_tuning       9.84 s      +36.06%     N/A         935.45 J    +35.17%     
# syrk_numpy             789.79 ms   +0.70%      N/A         80.95 J     +0.57%      
- syrk_omp               5.97 s      +31.05%     N/A         573.47 J    +29.89%     
- syrk_cuda              1.53 s      +12.48%     N/A         159.62 J    +10.93%     
- syrk_seq_tuning        5.98 s      +31.38%     N/A         574.51 J    +30.14%     
# trmm_numpy             880.51 ms   +0.39%      N/A         89.76 J     +0.50%      
# trmm_omp               711.70 ms   +0.01%      N/A         89.98 J     -0.63%      
# trmm_seq_tuning        3.38 s      +0.11%      N/A         276.75 J    -0.20%