LATX, AVX: write shuffle and byte-shift results directly

[luzeng87]

Summary

• allocate VPSLLDQ and VPSRLDQ destinations without loading overwritten values
• translate VPSHUFD with the direct LASX immediate shuffle
• write VPSHUFHW and VPSHUFLW directly to distinct destination registers while retaining a temporary for aliasing cases

Validation

• clean 3A6000 build: passed
• 128-bit-lane byte shifts and immediate dword/halfword shuffles: passed
• JIT, cold AOT, and hot AOT: passed with a non-empty AOT file
• git diff --check: passed

[luzeng87]

Superseded by grouped AVX lowering PR #448. The original commit remains separate inside that PR for review.