is_strided in layout_stride::mapping fails with extent 1 along a rank

Replicating layout_right as in test_layouts.cpp with layout_strided by setting
(extents 5,4,3,2,1, strides 24,6,2,1,1) is expected to generate a contiguous layout.

However, is_strided returns erroneously "false".

The issue arises because the sorting of strides with a permutation vector is now ambiguous at
https://github.com/ORNL/cpp-proposals-pub/blob/f8bf21b653d2424d40b9411ee532641f8c1c2e28/P0009/reference-implementation/include/experimental/bits/layouts.hpp#L272

In the following

https://github.com/ORNL/cpp-proposals-pub/blob/f8bf21b653d2424d40b9411ee532641f8c1c2e28/P0009/reference-implementation/include/experimental/bits/layouts.hpp#L285

This leads to the return of "false" because the wrong extent is used for comparison.

This only arises if one extent is 1, because only then, two strides would be the same.

Start with test_layouts.cpp and the layout_strided implementation in include/experimental/bits/layouts.hpp, especially the stride sorting around lines 272 and 285. Reproduce the extents and strides from the issue, then verify that is_strided reports the layout as contiguous when an extent is 1. Done means the regression case passes without changing behavior for other layouts.