Add deepsparse.analyze

[mgoin]

Adds a simple to run script that provides a simplified table to view the results of analyze_model(), so it is quick to view the performance profile of a model.

deepsparse.analyze zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned_quant-aggressive_95 -i "[1,128]" -q
Analyzing model: zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned_quant-aggressive_95
DeepSparse Engine, Copyright 2021-present / Neuralmagic, Inc. version: 0.12.0 (514d3953) (optimized) (system=avx512, binary=avx512)
== Layer Breakdown ==
Name                           | Summed Time | Percent Taken
Naive Subgraph 0               |    0.181    | 3.80%
elementwise                    |    0.232    | 4.88%
reduce                         |    0.573    | 12.04%
shuffle                        |    0.785    | 16.49%
ks_gemm                        |    1.975    | 41.50%
  kernel=[768, 768, 1, 1, 1]   |    0.703    | 14.76%
  kernel=[768, 3072, 1, 1, 1]  |    1.272    | 26.74%
elementwise_sub_2              |    0.009    | 0.18%
gemm                           |    0.283    | 5.95%
  kernel=[768, 2, 1, 1, 1]     |    0.012    | 0.26%
softmax                        |    0.108    | 2.28%
elementwise_plus_2             |    0.425    | 8.94%
Naive Subgraph 1               |    0.187    | 3.93%
== Summed Total Time: 4.7583 ms
== Items per second: 210.1591

[dhuangnm]

Thanks Michael.

@dbarbuzzi can you please add some tests for this utility after the PR lands? Thanks.