feat(opy): complete compiler directives and replacements

[e54-bot]

Summary

• Implement owner-side lowering and tooling support for issue Complete source-visible compiler directives and optimizer behavior #332.
• Keep translation lifecycle boundaries explicit; remaining translation work stays tracked by Complete OverPy translations and custom-game settings support #331.

Review fixes

• print("") now lowers the empty HUD header and color as Null, matching the pinned OverPy 9.7.10 semantic-WIR contract.
• #!debugElementCount now emits pinned-compatible counts: total 50, rules 32/18, conditions 3, and actions 14.
• The synthetic/debug-element-count fixture remains a semantic-WIR comparison and now also declares the CI-enforced debug-element-count source-visible projection from pinned workshopExact output; semantic WIR remains the secondary check.
• #!suppressWarnings filters matching preprocessing warning codes from tooling diagnostics.
• The compatibility-only semantic-WIR parser ignores upstream source-visible comments while preserving comment text in the public compiler output.
• Narrowed the top-level directive support wording and separated the remaining Complete OverPy translations and custom-game settings support #331 and unsupported directive boundaries.

Verification

• cargo fmt --all -- --check
• cargo test --locked -p opy-rs --all-targets --all-features — 414 passed, 1 ignored
• cargo clippy --locked -p opy-rs --all-targets --all-features -- -D warnings
• cargo clippy --locked -p opy-cli --features compatibility --all-targets -- -D warnings
• python3 -m unittest discover -s tools/overpy/tests — 28 passed
• Pinned synthetic/debug-element-count oracle snapshot: pass
• Full pinned OverPy 9.7.10 differential: 66 match, 10 known-gap, 0 regression across 76 fixtures; source-visible stage: 1 match

Independent ablation

• Removing the empty-print lowering correction regressed synthetic/directive-replacements semantic-WIR: 64 match, 10 known-gap, 1 regression.
• Restoring debug directive rejection failed the focused debug directive test with backend-directive-unsupported.
• Removing warning filtering failed the focused suppressWarnings test with a visible warning.
• Removing condition annotations failed the focused debug test: 3 emitted inline comments versus 5 expected.
• Removing summary sorting failed the focused debug test's descending-order assertion.
• Removing compatibility comment stripping failed the pinned debug fixture with malformed: expected a top-level section.
• Changing the pinned-compatible numeric/global-variable count from 2 to 1 caused the full differential to report synthetic/debug-element-count: source-visible regression (65 match, 1 regression).

Fixes #332

[Teakowa]

Blocking findings are inline. Also, #332 explicitly requires independent ablation of each substantive implementation change; the PR's Verification section currently lists only normal test/Clippy/differential runs. After fixing the blockers, record the requested ablation result so that acceptance criterion can be verified in the same follow-up review.

[e54-bot]

Review fixes are pushed in commit 2dba4a7. The pinned OverPy 9.7.10 differential now reports 65 match, 10 known-gap, 0 regression. Independent ablations recorded in the PR body show each substantive fix is necessary. Ready for re-review; hosted checks and reviewer approval remain separate.

[Teakowa]

Follow-up review: the replacement regression is fixed, the compatibility gate is green, #!suppressWarnings is implemented, and the requested ablation results are now recorded. One blocker remains from the original debugElementCount finding.

[e54-bot]

Follow-up fixes are pushed in commit 0878f9a. debugElementCount now annotates conditions and sorts the summary by count; the pinned regression fixture covers two differently-sized rules. The compatibility-only parser now ignores upstream comments so the fixture runs through semantic-WIR comparison. Full differential is 66 match, 10 known-gap, 0 regression across 76 fixtures. Ready for re-review.

[Teakowa]

Follow-up review: the implementation now covers condition/action comments and count-sorted summaries, and CI is green. One verification blocker remains: the new pinned fixture does not currently make the source-visible debug output part of the compiler gate.

[e54-bot]

Review fix pushed in commit 6bb149a. The pinned debugElementCount source-visible projection is now wired into the semantic-WIR fixture comparison, with semantic WIR retained as the secondary check. Full differential: 66 match, 10 known-gap, 0 regression across 76 fixtures; the new source-visible stage is 1 match. The numeric-count ablation produced a source-visible regression as expected. Local Rust tests, clippy, Python tests, and the pinned oracle fixture all pass; hosted CI and reviewer reapproval remain separate.

[Teakowa]

LGTM