[AArch64][SME2] Refine fcvtu/fcvts/scvtf/ucvtf (#77947)

Rename intrinsics for fcvtu to fcvtzu and fcvts to fcvtzs.

Use llvm_anyvector_ty for both multi vector returns and operands,
therefore the return and operands can be specified in the intrinsic
call, e.g.

@llvm.aarch64.sve.scvtf.x4.nxv4f32.nxv4i32