Reimplement rows case in zip_mut_with using Zip

[bluss]

• Reimplement the zip by rows case using Zip.
• Improves performance of the most important non-contiguous array shape: Where rows are contiguous but we have a greater stride from row to row. This appears if you split a contiguous array into rectangular pieces.

[bluss]

Ups and downs:

 name                       zipbefore.0 ns/iter  zipafter.0 ns/iter  diff ns/iter   diff % 
 add_1d_regular             450                  447                           -3   -0.67% 
 add_1d_strided             3,215                3,293                         78    2.43% 
 add_2d_0_to_2_iadd_scalar  353                  355                            2    0.57% 
 add_2d_alloc               979                  904                          -75   -7.66% 
 add_2d_assign_ops          467                  445                          -22   -4.71% 
 add_2d_broadcast_0_to_2    357                  353                           -4   -1.12% 
 add_2d_broadcast_1_to_2    641                  542                          -99  -15.44% 
 add_2d_cutout              916                  783                         -133  -14.52% 
 add_2d_cutouts_by_16       2,189                1,626                       -563  -25.72% 
 add_2d_cutouts_by_32       1,303                1,046                       -257  -19.72% 
 add_2d_cutouts_by_4        9,881                6,814                     -3,067  -31.04% 
 add_2d_f32_regular         452                  451                           -1   -0.22% 
 add_2d_regular             452                  443                           -9   -1.99% 
 add_2d_strided             2,155                2,593                        438   20.32% 
 add_2d_strided_dyn         6,696                3,058                     -3,638  -54.33% 
 add_2d_transposed          2,094                2,585                        491   23.45% 
 add_2d_zip                 444                  461                           17    3.83% 
 add_2d_zip_alloc           803                  840                           37    4.61% 
 add_2d_zip_cutout          902                  866                          -36   -3.99% 
 add_2d_zip_strided         2,563                2,578                         15    0.59% 
 add_2d_zip_transposed      2,549                2,587                         38    1.49% 
 assign_scalar_2d_corder    651                  649                           -2   -0.31% 
 assign_scalar_2d_cutout    857                  876                           19    2.22% 
 assign_scalar_2d_forder    653                  650                           -3   -0.46% 
 assign_zero_2d_corder      608                  626                           18    2.96% 
 assign_zero_2d_cutout      895                  875                          -20   -2.23% 
 assign_zero_2d_forder      625                  625                            0    0.00% 
 iadd_2d_regular            353                  355                            2    0.57% 
 iadd_2d_strided            1,568                1,600                         32    2.04% 
 scalar_add_1               745                  741                           -4   -0.54% 
 scalar_add_2               737                  743                            6    0.81% 
 scaled_add_2d_f32_regular  533                  539                            6    1.13%