Add provenance-linked real-project and minimized Workshop regression evidence

[Teakowa]

Parent: #10
Depends on: #18

Goal

Add real Workshop project evidence and provenance-linked minimized regressions so canonical correctness claims are grounded in externally meaningful behavior rather than only synthetic fixtures or implementation-local snapshots.

Context

workshop-rs already has focused parser/catalog/emitter/round-trip tests. #10 requires a distinct evidence layer where real project failures remain reproducible at both project and minimized-feature granularity, with explicit provenance and independently grounded expectations.

Scope

• Define the repository layout and metadata required for preserved real-project Workshop evidence and minimized regressions under the Define the Workshop conformance result contract and stable feature identities #18 provenance/result contract.
• Add at least one reviewed real-world Workshop project/revision/path as corpus evidence, respecting repository licensing/provenance policy.
• Identify at least one meaningful behavior or failure from real-project evidence and preserve a minimized reproduction linked back to the original project provenance.
• Keep the complete project-level case in addition to the minimized case so cross-rule, variable, settings, localization, and other integration interactions remain testable.
• Attribute regressions to stable Define the Workshop conformance result contract and stable feature identities #18 Workshop feature identities where applicable.
• Ensure expected behavior is independently supported by preserved source/client/reference evidence and is not established solely by snapshotting the current implementation output.
• Add corpus execution/reporting that distinguishes expected matches, known gaps, unsupported cases, regressions/divergence, and inconclusive evidence.
• Document how future real-project regressions should be admitted, minimized, reviewed, and retained.

Non-goals

• Replacing focused unit/property/census tests with real-project tests.
• Importing OPY or DEL/OSTW source projects as canonical Workshop semantics.
• Bulk-vendoring unreviewed community projects.
• Treating current workshop-rs output as an independent oracle.
• Requiring a live-client automation system before real-project evidence can land.

Acceptance criteria

• At least one real Workshop project case is preserved with repository/revision/path and licensing/provenance metadata.
• At least one project-derived behavior/failure has a minimized regression linked to the original evidence.
• Full-project and minimized cases remain separate complementary test layers.
• Conformance results use Define the Workshop conformance result contract and stable feature identities #18 states/feature identities and cannot report known failures as successful compatibility.
• The repository documents a repeatable evidence-admission/minimization process suitable for future regressions.