Complete the remaining Workshop-independent OPY syntax surface

[Teakowa]

Parent: #7

Goal

Close the remaining pure-language OPY syntax gaps that can be implemented and verified without Workshop lowering or catalog ownership.

Context

The current frontend already covers the declared core syntax/control-flow surface, but compatibility/support-matrix.json still records pure OPY syntax gaps such as switch/case/default, do…while, hex literals, expression-level membership, string modifiers, dicts, comprehensions, and lambda forms. These are language-frontend concerns and should be resolved before the Workshop integration stage.

Scope

• Implement and corpus-test switch / case / default.
• Implement and corpus-test do … while.
• Implement hexadecimal numeric literals where accepted by the pinned OverPy oracle.
• Implement expression-level in / not in semantics where independently evidenced.
• Implement the remaining inventory-backed string modifier forms (f/w/l/b/c/t) with reference-compatible parse/semantic behavior.
• Implement dict literals, list comprehensions, and lambda forms that belong to the declared OverPy compatibility surface.
• Preserve trivia/source spans/provenance and recoverable diagnostics so tooling APIs remain usable on the new constructs.
• Add positive/negative minimized fixtures and oracle probes for every newly supported form.
• Update human and machine-readable compatibility evidence only when implementation/tests justify the status change.

Non-goals

• Workshop lowering or emission.
• Canonical Workshop builtin/enum/member validation.
• Inventing WrightKit-only OPY syntax or improving upstream syntax beyond observable compatibility.
• Matching upstream parser internals or formatting.

Acceptance criteria

• The pure syntax rows currently tracked as planned no longer remain planned without an explicit evidence-backed reason.
• Reference-accepted syntax in scope parses/resolves through Opy HIR or an appropriate OPY semantic representation with stable source provenance.
• Invalid/reference-rejected variants produce structured diagnostics rather than panics or silent acceptance.
• Differential evidence distinguishes genuine native gaps from unsupported/out-of-scope behavior.
• Existing tooling APIs and compatibility corpus remain green.
• No workshop-rs dependency or Workshop catalog data is introduced.