Remove the wright-workshop cutover adapter (complete the wright#143 removal path)

[Teakowa]

Summary

Completes the entropy-audit candidate with the strongest evidence: wright-workshop was a 17-line re-export-only cutover adapter (pub use workshop_rs::*;) whose own doc comment prescribed this exact removal — migrate call sites to workshop_rs::, then delete the crate. The v0.2 cutover is done (workshop-rs is at released 0.1.5); this PR executes the documented follow-through.

• Call-site migration: all 111 wright_workshop:: references in wright-analyzer, wright-opy, wright-driver, wright-ostw (src and tests) now use the identical workshop_rs:: paths. All four crates already declared workshop-rs as a direct dependency, so this is mechanical; several files had been mixing both paths since the cutover.
• Preserved regressions: the ten [OPY baseline] Implement semantic compatibility manifest and systematic builtin/member/signature coverage #109/[Workshop semantics] Resolve ambiguous enum members using expected parameter domains #111 manifest-wiring tests — the shipped driver's ChainedExpectedDomain parse context, the chase/Set Invisible bare-None fixed points (including the negative context-free case), the manifest expected-domain contract, and the chase-keywords fixture end-to-end OPY→HIR→WIR→emit→reparse round-trip — move to crates/wright-driver/tests/manifest_context.rs, colocated with the CompilerSession wiring they protect (session.rs parse path).
• Deleted coverage: the adapter's remaining ~2.5k test lines exercised workshop-rs's own catalog/parser/emitter/round-trip/detect/closure behavior through the re-export. workshop-rs owns and runs that coverage itself (crates/workshop-rs/tests/, 18 suites including all eight counterpart files), so these were duplicate maintenance, not Wright integration evidence.
• Docs/comments: 17 documents and 4 code comments described consumption "via the wright-workshop adapter"; they now state direct workshop-rs consumption. ADR-0009's migration-state list records the adapter as removed (wright-opy/wright-ostw/wright-ir remain pending their v0.3/v0.4 extractions).

Why this is safe

• The adapter contained no implementation; every re-exported path is the workshop-rs path, so behavior is unchanged by construction.
• No public/published surface changes: no crate in this workspace is published to crates.io; releases ship only the wright and wright-lsp binaries.
• The deleted tests' subject matter has an owning test suite in workshop-rs (counterpart files verified: catalog, closure, detect, emitter, parser, pipeline, roundtrip, wir_expansion).

Validation

From the branch (all commands run locally, will be re-proven by CI):

• cargo fmt --all -- --check — clean
• cargo clippy --locked --workspace --all-targets --all-features -- -D warnings — clean
• cargo test --locked --workspace --all-targets --all-features — 591 passed / 0 failed (includes the 10 preserved tests: cargo test -p wright-driver --test manifest_context → 10 passed)
• grep -rn "wright.workshop" across rs/toml/md — only 4 intentional historical notes remain (test provenance comment, ADR-0009 status, compatibility.md and catalog-pipeline.md removal records)
• git diff --check — clean

Testing-policy notes

• Expectations were not changed to match a new implementation; behavior is identical by re-export equivalence, and all surviving observable-contract tests (including the preserved [Workshop semantics] Resolve ambiguous enum members using expected parameter domains #111 negative paths) still pass unmodified apart from import paths.
• The moved tests keep their original assertions, comments, and issue linkage ([OPY baseline] Implement semantic compatibility manifest and systematic builtin/member/signature coverage #109/[OPY baseline] Implement named/keyword argument binding and contextual call semantics #110/[Workshop semantics] Resolve ambiguous enum members using expected parameter domains #111) for review provenance.

Net effect

46 files, +383 / −2769. Removes one workspace crate concept, one dependency edge from each of four crates, eight duplicate test suites, and the dual-path (wright_workshop::/workshop_rs::) cognitive overhead. No capability lost; wright-ir's four ~50-line cutover shims remain, tied to the v0.3 HIR extraction per their own doc comments.

[Teakowa]

Force-pushed a rebuilt branch: the original push had accidentally been based on a local main that contained the unmerged feat/issue-213-terminal-presentation commit (4f67f98\#216), so this PR's diff and CI run included that PR's wright-cli presentation changes and inherited its CI-only terminal-coloring test failure (terminal_renderer_uses_command_specific_hierarchy fails under CI=true/GITHUB_ACTIONS=true where colors are emitted without a TTY — needs fixing in #216, not here).

The branch is now exactly origin/main (ac8d14f) + the 46-file adapter removal, cherry-picked with no wright-cli files. Revalidated on the clean base: cargo fmt --check clean, cargo clippy --workspace --all-targets --all-features -D warnings clean, cargo test --workspace 588 passed / 0 failed.