[Review] Gennaro Condition 1

[mborland]

Resolve the UB in strided/transposed end() iterators

[correaa]

I would like to request a temporary pass on this one. Please grant it, as this issue is the hardest problem in the review and, in practice, doesn't yet affect usage. I take UB seriously, and I understand the problem.

Instead of this resolution, please consider this clear documentation: https://correaa.gitlab.io/boost-multi/multi/technical.html#technical_iteration. I promise to keep exhaustive tests on as many platforms as I can realistically access, even beyond the normal Boost CI.

If the library is accepted, I will prioritize this in the near future. The mitigation strategies are ranked here: https://correaa.gitlab.io/boost-multi/multi/technical.html#technical_library_level_mitigations, and the solution will involve improving the user experience as the user base grows.

The current line of work is point 1 in https://correaa.gitlab.io/boost-multi/multi/technical.html#technical_library_level_mitigations
(use internal counters to avoid pointer representation problems)
Also to "show work", I have this experimental branch that will require some time and testing that it is not disruptive: https://gitlab.com/correaa/boost-multi/-/merge_requests/1919