test(sync): add minimal A5 TMOV col-major alignment repro

[TaoTao-real]

Summary

• Add a minimal A5 repro sample for TMOV on 16x1 f32 col_major tile in test/samples/Sync/test_tmov_col_major_16x1_align_a5.{py,pto}.
• Add a row-major control sample in test/samples/Sync/test_tmov_row_major_1x16_control_a5.{py,pto}.
• Extend test/samples/runop.sh:
  • Skip these two samples when --pto-arch is not a5.
  • Add guards to ensure generated C++ keeps TMOV and expected tile layout (ColMajor 16x1 / RowMajor 1x16).

Motivation

We need a focused TMOV-only pair to validate whether A5 runtime failures are caused by TMOV_V2V alignment behavior for col_major (RowStride=1) shapes.

Validation

• bash -n test/samples/runop.sh
• Generated and compiled both new samples with A5:
  • ptoas --pto-arch=a5 --enable-insert-sync .../test_tmov_col_major_16x1_align_a5.pto
  • ptoas --pto-arch=a5 --enable-insert-sync .../test_tmov_row_major_1x16_control_a5.pto
• Verified emitted C++ contains:
  • Repro: TMOV + Tile<..., 16, 1, BLayout::ColMajor, ...>
  • Control: TMOV + Tile<..., 1, 16, BLayout::RowMajor, ...>

Notes

This PR adds compiler/sample coverage only. Runtime pass/fail signal should be collected on A5 board by running these two generated kernels side-by-side.