Should gsl::span::index_type be harmonized with std::span::index_type

[JordanMaples]

gsl::span is std::ptrdiff_t whereas the std::span is std::size_t. Should we update the guidelines to have these definitions converge?

[hsutter]

Editors' call: Normally when a GSL type is standardized we drop the GSL type. In this case, we're not withdrawing GSL span because some users rely on checking, which std::span doesn't do. But we should make gsl::span as close as possible to std::span, including adding CTAD and considering changing signed to unsigned for index_type.

See also #1497 and #1504.

[hsutter]

Editors' call: We feel it's time to switch to unsigned. We still feel that using signed is the better solution, but the standards committee has decided on unsigned, and we feel we should not diverge from the standard on this point. (In contrast, we do feel we must diverge for guaranteed checking.)