TODO: Conformance fixtures and cross-language parity

Goal

Grow shared IR fixtures with expected Julia codegen (and later expected trajectories or numerical checks), executed in Python and Julia (ModelParserJL) so the two runtimes cannot drift silently.

Why

• docs/design/language-strategy.md: the IR file is the boundary; parity tests are the enforcement mechanism.
• Reduces regression risk whenever MTK idioms or codegen change (see ADR 0004).

Scope hints

• Start with golden .jl snippets or full files under tests/ or examples/.
• Julia side: documented Pkg.test or script in CI (may remain optional job until Julia CI is wired).

References

• docs/design/model-parser.md roadmap §2.
• ADR 0001, ADR 0004.

Acceptance sketch

• At least N≥2 fixtures with checked emit julia output.
• Document how to refresh goldens when codegen intentionally changes.