[clang] Honor pragmas with -ffp-contract=fast, depecate fast-honor-pragmas (PR #105746)

[Andy Kaylor via cfe-commits]

andykaylor wrote:

This may not be as bad as I thought. After looking more closely at the one test case that was failing with this change, it turned out that the failure was happening because the test contained a pragma to turn contraction off but checks to verify that the fma was generated anyway. Unless that behavior is needed, this may be OK.

https://github.com/llvm/llvm-project/pull/105746