[AMDGPU] Make the options consistent across 3 RA pipelines(NFC) (#184190)
Adding the missing option for the wwm-regalloc in the test
attr-amdgpu-flat-work-group-size-vgpr-limit.ll. The existing
test already specifies -sgpr-regalloc=fast & -vgpr-regalloc=fast
to ensure that the fast register allocator is preferred over
the default greedy allocator. For consistency, the same
preference should also be applied to the wwm-regalloc pipeline.