Implement ranges::starts_with and ranges::ends_with

[miscco]

This implements P1659 from @cjdb

I really wish we had something like an constexpr if-expression

Fixes #1975.

[miscco]

Note I did not yet add the feature macro, what is the right value?

[miscco]

@CaseyCarter it seems that mismatch was one of the first ranges algorithms so it has some oddities. I think we should move the unwrapping out of the helper functions and into the main function calls as in the other algorithms.

If so should that happen in this PR?

[miscco]

I believe we should clean up _Mismatch_fn because currently we have a mismatch whether the algorithms expect uwrapped or plain iterators

[miscco]

@CaseyCarter, do you have some insights on what to do if the needle range is infinite?

Should that return true or false?

[miscco]

@AdamBucior I added the requested optimizations

[CaseyCarter]

I really wish we had something like an constexpr if-expression

You mean something other than if constexpr and if consteval? I suspect EWG will be reluctant to add more variations of compiletime if.

Note I did not yet add the feature macro, what is the right value?

202106L

Because of my old Nemesis. Input iterators cannot be copied into distance

I'm working on #1684 now.

@CaseyCarter it seems that mismatch was one of the first ranges algorithms so it has some oddities. I think we should move the unwrapping out of the helper functions and into the main function calls as in the other algorithms.

I vaguely recall this being useful for some reason. Feel free to take a stab at it, and if something goes pearshaped we can add an explanatory comment so future us doesn't try to fix this again.

If so should that happen in this PR?

This PR won't be ported to 16.11, so I wouldn't complain about unrelated cleanup - your call.

I believe we should clean up _Mismatch_fn because currently we have a mismatch whether the algorithms expect uwrapped or plain iterators

Agree.

@CaseyCarter, do you have some insights on what to do if the needle range is infinite? Should that return true or false?

It should return false if a mismatch is found and otherwise run forever.

[StephanTLavavej]

This PR was previously passing checks, but is now failing after the MegaMerge, with apparently-real failures with both MSVC and Clang in different tests.

[CaseyCarter]

@miscco FYI: I thought we could do better in ends_with than to iterate over both ranges twice when non-sized, so I implemented some special cases. (I may have gone overboard.)

[CaseyCarter]

Chained comparisons don't really work this way in C++ - shout if people think this comment is unclear.

[StephanTLavavej]

I noticed it but decided not to meow. 🐱

[miscco]

Thanks for going the extra mile @CaseyCarter

Looks good to me

[StephanTLavavej]

I'm mirroring this to an MSVC-internal PR; please notify me if any further changes are pushed.

[StephanTLavavej]

Thanks for implementing these new ranges algorithms! 🚀 😻 🛠️