Establish the canonical multi-locale Workshop core

[Teakowa]

Depends on: #1
Wright cutover: wright#143

Goal

Build workshop-rs as the separately versioned canonical Workshop semantic core and standalone Workshop tooling foundation for WrightKit.

Scope

• Bootstrap the repository with its package/workspace, CI, tests, README, license/provenance policy, and repo-local AGENTS.md.
• Extract canonical Workshop ownership from Wright without importing Wright tooling responsibilities.
• Own locale-independent Workshop semantic IDs, WIR/canonical semantic representation, parser/CST, validation, emitter, and canonical catalog integration.
• Complete raw Workshop parsing/validation/emission for the declared supported surface, including settings.
• Add localized raw Workshop input/output and first-class raw Workshop locale conversion.
• First release gate: corpus-tested en-US <-> zh-CN conversion for the declared surface, including settings and representative actions/values/events/enums/content identifiers.
• Preserve user-authored identifiers, strings, and comments according to a documented policy; localize Blizzard-defined Workshop language/content tokens only by default.
• Missing target-locale mappings fail explicitly by default; any fallback is opt-in.
• Provide standalone library and CLI surfaces suitable for use without Wright.

Non-goals

• OPY or DEL/OSTW-compatible source-language semantics.
• Wright lint/agent/orchestration behavior.
• Guaranteeing every Overwatch client locale in the first release.
• Guessing unsupported future syntax.

Acceptance criteria

• workshop-rs has no dependency on Wright tooling crates or source-language provider internals.
• Canonical Workshop IDs and semantic representation are locale-independent.
• The supported raw Workshop surface round-trips through parse -> canonical semantics -> emit.
• en-US -> zh-CN and zh-CN -> en-US conversion pass the declared corpus, including settings.
• Catalog provenance/version identity is machine-readable and reproducible.
• CLI/library usage is documented and tested.
• The repository exports a stable enough dependency surface for Wright to cut over without keeping a second canonical Workshop implementation.

[Teakowa]

Wright integration/cutover is tracked separately in wrightkit/wright#143; keep this issue focused on the standalone Workshop core.

[Teakowa]

v0.2 status: canonical core established and merged (PR #4, commit b96896f) and consumed by wright (wright#143 cutover merged). Acceptance criteria 1-3 and 5-7 MET with QA evidence (124 tests, CI green). Criterion 4 (corpus-tested en-US<->zh-CN conversion for the declared surface) is the remaining v0.2 gate: zh-CN is declared with 0/344 mappings; fail-explicit MissingMapping, opt-in fallback, catalog identity/digest, and the locale mechanism are proven (incl. synthetic-locale bidirectional conversion). Re-scoped follow-up: add an evidence-backed zh-CN mapping corpus once a permissible, provenance-reviewed source passes ADR-0001 (OverPy GPL tables excluded). This issue stays open as the zh-CN corpus gate.

[Teakowa]

zh-CN corpus gate: user has chosen to provide permissible zh-CN Workshop evidence (client snapshots/spellings) for provenance-reviewable transcription per ADR-0001. Once the evidence location is available, the corpus work proceeds: add zh-CN aliases with provenance records via the catalog pipeline (workshop-catalog-gen check/build), then demonstrate corpus-tested bidirectional en-US<->zh-CN conversion incl. settings to satisfy the v0.2 gate.

[Teakowa]

Current v0.2 status

The remaining zh-CN lane is represented by Draft PR #9, now at 38ba22c.

Evidence currently landed:

• catalog coverage: 327/344 canonical entries; all 17 exclusions remain fail-explicit exact-match exclusions
• settings coverage: labels 17/19, modes 7/7, maps 2/2, heroes 10/10, enums 2/2, tokens 3/3, teams 1/1
• corpus source commit: d854bf01fc7bbf3b2169f67408c07a8da8989ad6
• export SHA-256: e6a80867407d9c432dbbafe43b0d7121b9fb2de969eff8f15be114bf18e1a717
• CI: stable and Rust 1.85.0 both pass

The full release gate is not claimed. The remaining blockers are the 17 unmapped catalog identities, two Blizzard hero-settings templates absent as exact names from the export, and final redistribution/license review. For Player Variable also remains outside the declared catalog surface.

Next dependency: review/approve permissible provenance and provide exact evidence for the remaining declared spellings, or explicitly accept a revised declared surface. No approximate or inferred mappings were added.