Remove workaround from floatdouble and floatsingle after the /fp:fast bugs are resolved

[tannergooding]

dotnet/coreclr#19725 added workarounds to floatdouble.cpp and floatsingle.cpp to ensure that ceil and floor return the appropriate values when compiled with MSVC.

This is tracked internally in the DevDiv bug database as workitem 673062 and 673060.

[AaronRobinsonMSFT]

Talking to internal teams it appears the current POR is to have these delivered in a similar time for the release of 3.0. This means there probably isn't time to remove the workarounds prior to those fixes. Moving this to 3.x timeframe.

[tannergooding]

I also don't believe we can remove these until we make the new CRT a baseline requirement for running CoreCLR (I don't believe we want someone hitting this issue if they build locally using VS 2017, for example).

We can certainly change the #ifdef, however 😄

[mangod9]

@tannergooding @AaronRobinsonMSFT triaging through old .net 5 issues. Is this something which is critical for 5?

[tannergooding]

It isn't and isn't something we can fix either; it is a bug in the MSVC compiler when fp:fast is specified.

[mangod9]

ok thanks for the info. Should it be moved to a different area than vm-coreclr? Will move to 6 for now.