feat(opy-compiler): lower OPY structure into canonical WIR

[Teakowa]

Summary

• implement Lower OPY program structure into canonical Workshop WIR #40 structural HIR to canonical workshop-rs WIR lowering for variables, subroutines, rules, events, filters, and provenance
• reserve explicit indices before deterministic allocation and preserve source/name spans
• keep unsupported settings, metadata, declarations, events, filters, and statement forms explicit and source-attributed
• update the support matrix to separate the Lower OPY program structure into canonical Workshop WIR #40 structural slice from broader lowering-dependent Workshop work

Scope

• Fixes Lower OPY program structure into canonical Workshop WIR #40
• Does not implement Lower OPY statements and expressions into canonical Workshop WIR #41, Close catalog-backed OPY lowering and corpus evidence #42, Integrate Workshop-dependent settings, locale, directive, and hook behavior #37, or full Integrate OPY lowering and end-to-end compatibility with workshop-rs #8 Workshop lowering
• No copied Workshop catalog data or temporary IR

Checks

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• cargo test --workspace --all-targets --all-features
• python3 -m unittest discover -s compatibility/tests

[Teakowa]

PM/architecture audit: #44 is structurally close, CI is green, but I would not merge this revision yet.

Required fixes:

1. Preserve subroutine source/runtime identity across @Name and rule-prefix presentation metadata. The frontend intentionally carries both SubroutineDef.name (generated/presentation rule name) and source_name (the source subroutine symbol). lower_subroutine() currently rewrites WorkshopSubroutine.name to the presentation name and aliases that name into the semantic subroutine map. This contradicts the pinned reference: subroutine helper + @Name "renamed helper" emits subroutines { 0: helper } and the subroutine event target helper, while only the generated rule title becomes [Source] renamed helper. Keep the WIR subroutine identity/name as source_name; use the generated name only for the WIR rule presentation name. Update the structural test accordingly.

2. Do not make @SuppressWarnings a Workshop-lowering failure. It is OPY/frontend diagnostic metadata and has no runtime Workshop carrier to emit. The pinned directive fixture compiles successfully with @SuppressWarnings on the subroutine/rule and emits no corresponding Workshop construct. Treat it as already-consumed tooling metadata (preserve it in HIR if needed), rather than unsupported-integration-surface. @Delimiter/@NewPage may remain explicit gaps because they do affect generated Workshop structure/presentation.

3. Add independent Lower OPY program structure into canonical Workshop WIR #40 compatibility evidence before marking the structural slice end-to-end-supported. This PR changes only compiler code/docs; its new structural tests self-assert implementation behavior, while the support-matrix row cites the Establish the versioned OPY-to-Workshop integration boundary #35 fixture that does not exercise subroutine identity, explicit/implicit index allocation, or player event filters. Add minimized oracle-backed Lower OPY program structure into canonical Workshop WIR #40 fixture(s) covering at least subroutine @Name identity, deterministic explicit/implicit allocation, and event filters, then use those snapshots plus WIR assertions as evidence. Do not advance the support row solely from implementation-local tests.

What is already accepted:

• dedicated opy-compiler boundary remains intact;
• explicit-index reservation/collision diagnostics are structured;
• subroutine/rule/event WIR nodes preserve spans;
• unsupported structural forms fail explicitly instead of being silently dropped;
• no temporary Workshop IR or copied catalog tables were introduced.

After these corrections, rerun full workspace CI + compatibility/oracle checks; #40 should then be ready for final acceptance.

[Teakowa]

Final review: the three prior semantic/evidence blockers are resolved. One testing-policy cleanup remains before merge.

compatibility/tests/test_runner.py changes the exact fixture-count assertion from 41 to 42. WrightKit testing policy explicitly avoids hard-coded corpus-count assertions as a proxy for completeness; adding a fixture should not require incrementing a magic number. Please replace/remove the exact-count assertion and keep the meaningful invariants instead (fixture discovery succeeds, metadata/snapshots are valid, declared expectations cover discovered fixtures, no unexpected divergence/skips as applicable).

This is not a #40 semantic blocker; it is a merge-quality/testing-policy cleanup. After that and green CI, I consider #40 ready.