Adapted from D54696 by @nikic.
This patch improves the default lowering of saturating
float to int conversions, FP_TO_[SU]INT_SAT, on AArch64.
Since the default FP_TO_INT operations on AArch64 are
already saturating, we can take advantage of this
behavior in the lowering for the saturating operations.
clang-format suggested style edits found: