LATX, AVX: defer redundant VEX.128 upper-half clears

[luzeng87]

Summary / 变更说明

• Track VEX.128 destinations whose architectural YMM high halves are known to be zero.
• Defer redundant LASX high-half clears until a 256-bit operation can observe the state or the TB is about to exit.
• Materialize pending clears before guest-visible exits so JIT, TU, AOT, and signal state keep x86 semantics.
• Add a standalone semantic test covering native x86, LAT JIT, cold AOT, and hot AOT execution.

This change does not modify the AOT cache identity or enable LAT --optimize-O2.

Validation / 验证

• 3A6000-25G, GCC 8.3.0 -O2 -g, LAT --optimize-O1: clean x86_64 build passed.
• meson test -C build64-pr --suite lat-pr-fast: 24 passed, 0 failed.
• vex128-ymmh-zero.S: native x86, LAT JIT, cold AOT, and hot AOT all passed; non-empty hot cache was loaded.
• The stacked GB603 branch containing this commit completed a clean i386 LAT build (407/407 build steps).

Checklist / 检查项

• I have read CONTRIBUTING.md. / 我已阅读 CONTRIBUTING.md。
• Every commit contains a DCO sign-off (git commit -s). / 每个提交都包含 DCO 签署。
• I have included relevant build or test results. / 我已提供相关构建和测试结果。