Implement ranges::replace

[miscco]

This implements the ranges::replace algorithm.

@CaseyCarter: This does not yet compile as the algorithm has some additional constraints that are not fulfilled by the generic with_input_range

The question is whether we want to specialize some with_comparable_range in the algorithm support header or simply special case it in the test. As that will most likely happen with other algorithms too I would like find a fitting solution that fits best for all algorithms.

Also the wording is really strange. IT says that the algorithm shall return last which could actually be a sentinel. I believe it is indeed meant to return respective iterator but icky wording

[CaseyCarter]

This does not yet compile as the algorithm has some additional constraints that are not fulfilled by the generic with_input_range

Since this is passing now, I suspect you've realized that testing ranges with proxy references requires you to stick fairly closely to ints or pairs, or be prepared to add operations to test::proxy_reference.

also the wording is really strange. It says that the algorithm shall return last which could actually be a sentinel. I believe it is indeed meant to return respective iterator but icky wording.

See [algorithms.requirements]/13, and read [algorithms.requirements]/12 while you're there. WG21 specifies algorithms in a DSL that resembles but is not C++ ;)

[miscco]

One style question for the maintainers.

Do you prefer individual algorithm PRs like this here or should I open a single PR for e.g. all ranges::replace* algorithms?

[StephanTLavavej]

I think it's generally easier to review an entire family of algorithms in a single PR, since following a consistent pattern makes reviewing similar algorithms easier - then we just need to watch out for inconsistencies and (somewhat harder) watch out for places that should be different - neither of which is helped by splitting algorithms across PRs.

Other maintainers may feel differently. It's not critically important for me.

[miscco]

To ease review and reduce thenumber of concurrent PRs I have merged the other two remaining ranges::replace algorithms into this PR

[StephanTLavavej]

Will push changes to fix comment typos, otherwise looks great!

[StephanTLavavej]

Thanks @miscco! You're irreplaceable. 😎