lp64s is same as lp64d execpt that floating point arguments and return
values are always passed via GPRs or stack which means UseGPRForFloat
is always true in CC_LoongArch for lp64s.
One motivation of this change is to build linux which uses
-msoft-float and -mabi=lp64s [1].
It sounds like we can avoid having to rename the file and change this comment one more time when we wire up LP64F, by rewording the sentence a little bit and renaming the file? Like "calling-conv-common.ll" (leaving room for e.g. "calling-conv-fp-common" or "calling-conv-vector-common") and "... identical output for all ABIs, i.e. ...".