Add .as_slice_memory_order(), improve scalar_sum, and fix bugs in from_vec_dim_stride

[bluss]

• Add .as_slice_memory_order(), .as_slice_memory_order_mut()
• Fix bugs in from_vec_dim_stride that relate to axes of length 0 or 1.
• Misc bonus changes: Add Send + Sync to NdFloat, Improve Display for ShapeError.

Part of #138
Fixes #58
Fixes #144

Using cargo-benchcmp to review performance changes

The true improvement here is assign scalar for the transposed (f-order) case and scalar_sum should improve the same way. Most important was though to validate that there are no perf regressions. Everything looks like it's the same or within error.

name                               assign_scalar_before2 ns/iter  assign_scalar_after2 ns/iter    diff ns/iter   diff %
assign_scalar_2d_cutout            1,640                          1,687                                     47    2.87%
assign_scalar_2d_large             1,595                          1,593                                     -2   -0.13%
assign_scalar_2d_transposed_large  3,618                          1,644                                 -1,974  -54.56%

name                       add_before3 ns/iter  add_after2 ns/iter    diff ns/iter   diff %
add_2d_0_to_2_iadd_scalar  476                  399                            -77  -16.18%
add_2d_assign_ops          626                  616                            -10   -1.60%
add_2d_broadcast_0_to_2    476                  400                            -76  -15.97%
add_2d_broadcast_1_to_2    646                  652                              6    0.93%
add_2d_cutout              1,254                1,257                            3    0.24%
add_2d_f32_regular         632                  637                              5    0.79%
add_2d_regular             625                  628                              3    0.48%
add_2d_transposed          3,341                2,906                         -435  -13.02%
muladd_2d_f32_regular      699                  700                              1    0.14%

name                        scalar_sum_before4 ns/iter  scalar_sum_after5 ns/iter    diff ns/iter  diff %
scalar_sum_2d_cutout        745                         732                                   -13  -1.74%
scalar_sum_2d_float         570                         578                                     8   1.40%
scalar_sum_2d_float_cutout  839                         838                                    -1  -0.12%
scalar_sum_2d_regular       251                         253                                     2   0.80%

[bluss]

@SuperFluffy the ordering agnostic version of as_slice() makes its appearance here.