[stdlib] fix another accidental infinite-recursion bug

[glessard]

It's been known for years (see SR-6501) that it's easy to forget to implement the critical function replaceSubrange<C>(_: with: C) when conforming a type to RangeReplaceableCollection, because the compiler doesn't ask for it. At runtime, there is then an infinite-recursion crash on just about any RRC-specific function.

This adds an unavailable implementation to transform this issue into a compile-time error, in line with other recent similar fixes, such as #36969.

Resolves SR-6501

[glessard]

@swift-ci please test macOS platform

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 8051a46d8c6681d939ba51f6e26f726c136cf761

[glessard]

@swift-ci please test macOS platform

[glessard]

@swift-ci please smoke test

[glessard]

@swift-ci Please Test Source Compatibility

[glessard]

@swift-ci Please Test Source Compatibility

[glessard]

@swift-ci Please Test Source Compatibility

[stephentyrone]

LGTM

[dduan]

fwiw this broke source compatibility. For a good reason, ofc!

[glessard]

Yeah, breaking code that relies on a bug is deemed not to be an actual source break.

[stephentyrone]

Expanding a little on what Guillaume said, code that was "broken" by this change merely compiled; it never actually worked because it had an invalid conformance to RangeReplaceableCollection. Thus, there is no source break--conformances must satisfy the semantic contract of the protocol in a Swift program*.

(*) Yes, the program used to compile, but it was not a valid Swift program; we simply were unable to detect the constraint violation at compile time. Now we can.