NOT READY FOR MERGE: add attention looping verification harness

[booth-algo]

Summary

Adds testbench/report artifacts for ATen MHA attention-looping work:

• Adds attention_looping_compare.py for looped-vs-unrolled static/dynamic/cycle comparison.
• Adds flash_attention_mha_test.py to run ATen per-head MHA through the transactional emulator against PyTorch SDPA golden.
• Adds doc/ATTENTION_LOOPING_STEP2_REPORT.md with commands, branch hashes, estimates, and golden-check output.

The actual compiler implementation is in the companion PLENA_Compiler PR from feat/codegen-addr-reg-init.

CLM-60M Native Layer 0 Counts

Rerun locally with compile_hf_model(model, seq_len=64, hidden_size=None, inter_dim=None, num_layers=1). Native dims: hidden=384, inter=1408, heads=6, kv_heads=2, head_dim=64.

Metric	Previous	Current	Change
Total ASM source lines	35,479	15,367	-20,112 (-56.7%)
Actual static instruction lines	34,041	14,403	-19,638 (-57.7%, 2.36x smaller)
Comment / metadata lines	1,438	964	-474 (-33.0%)
Loop-expanded dynamic instructions	645,334	649,762	+4,428 (+0.69%)
Estimated cycles	8,915,366	8,919,794	+4,428 (+0.05%)
Estimated ms @ 1GHz	8.915366	8.919794	+0.004428 (+0.05%)
C_LOOP_START static lines	248	296	+48

Previous file analyzed: /tmp/clm60m_native_1layer/generated_asm_code.asm. Current file analyzed: /tmp/clm60m_native_1layer_after_attention_looping/generated_asm_code.asm.

Verification

• py_compile for attention helper and harness files
• attention_looping_compare.py for head_dim=64
• attention_looping_compare.py for head_dim=128
• attention_codegen_compare.py after attention looping
• flash_attention_mha_test.py transactional emulator golden check

Status

Not ready for merge. Branch has been replayed onto current main; opened for review/context while the related compiler branch is still being validated.