cleanup of FPE

[rrsettgast]

This PR seeks to cleanup our treatment of FPE and underflow flushing. Goals:

• cleanup platform branching logic
• ensure testing is working on all platforms
• ensure underflow flush to zero is properly enabled

[wrtobin]

rocmcc supports the cuda flag: -fcuda-flush-denormals-to-zero

https://rocm.docs.amd.com/en/docs-6.0.2/reference/rocmcc.html

[wrtobin]

misclick