Some float controls fixes for lower_doubles
The fp64 fixes from !27281 (merged) The lower_doubles pass was not respecting float_controls flags, as well as a couple fixes to nir_algebraic to avoid double lowering some of these instructions.
The fp64 fixes from !27281 (merged) The lower_doubles pass was not respecting float_controls flags, as well as a couple fixes to nir_algebraic to avoid double lowering some of these instructions.