Share implementation between std::atomic::wait for P1135R6 (#52) and parallel_algorithms.cpp

[AlexGuteniev]

parallel_algorithms.cpp uses specific case of wait-on-address, which waits on one byte.

As Billy O'Neal (@BillyONeal) pointed out, it makes sense to switch it to generic implementation, when such generic implementation is ready.

Even though new generic implementation would be almost identical (except that it would handle odd atomic sizes), it makes sense to switch to generic one to avoid duplication.

[StephanTLavavej]

My understanding is that because parallel_algorithms.cpp is injected into the import lib, changes here wouldn't affect binary compatibility, so this doesn't need to be marked as a vNext issue (which is good news).

[BillyONeal]

In fact, the reason the import lib trick worked for parallel_algorithms.cpp is that we know execution never leaves the module (that's why I was able to get away with avoiding a satellite DLL for the parallel algorithms' missing-WaitOnAddress-fallback-lock-table).

[AlexGuteniev]

How should code be shared? By linking to atomic wait satellite dll/lib?

Have an idea, just add satellite.cpp to ordinary build. It will not export DLL symbols, as there's no .def file.

[AlexGuteniev]

Incorporated this change into #593 , tests pass.