ANS-1121: add optimizer judge model flag

[jharris1679]

Summary

• add --judge-model as an explicit optimizer role override
• make shared --model apply to eval, judge, trace analysis, and component author roles
• preserve each eval case configured judge when neither shared nor explicit judge model is supplied
• document the role behavior and add CLI request-contract coverage

Validation

• npm test — 72 passed
• git diff --check — passed

Companion to the answerlayer-core ANS-1121 implementation.

[greptile-apps]

Greptile Summary

Adds optimizer model-role selection and documents its fallback behavior.

• Adds --judge-model as an explicit judge override.
• Applies the shared --model to eval, judge, trace-analysis, and proposal roles.
• Preserves case-configured judges when neither a shared nor explicit judge model is supplied.
• Adds request-contract coverage for explicit, shared, and omitted judge models.

Confidence Score: 5/5

The PR appears safe to merge with the model-role behavior correctly implemented, tested, and documented.

No actionable new issues remain, and the previous documentation finding is fully addressed by explicitly describing the analysis-model fallback order.

Important Files Changed

Filename	Overview
src/cli.js	Adds parsing, help text, and request serialization for the judge-model override.
test/cli.test.js	Verifies explicit, shared, and omitted judge-model request behavior.
README.md	Documents role overrides and the proposal-to-analysis fallback, resolving the previous finding.

_{Reviews (2): Last reviewed commit: "docs: clarify optimizer model precedence" | Re-trigger Greptile}

[greptile-apps]

Analysis fallback is undocumented

The role description presents --proposal-model and --analysis-model as independent overrides, but the CLI resolves the analysis model from --analysis-model, then --proposal-model, then the shared --model. Without this detail, users may unintentionally select their proposal model for trace analysis.

Suggested change

	The shared `--model` defaults the eval agent, judge, trace analyzer, and
	component author. Use `--eval-model`, `--judge-model`, `--analysis-model`, or
	`--proposal-model` for explicit role overrides. Without a shared model or
	`--judge-model`, each case keeps its configured judge.
	The shared `--model` defaults the eval agent, judge, trace analyzer, and
	component author. Use `--eval-model`, `--judge-model`, `--analysis-model`, or
	`--proposal-model` for explicit role overrides. Without `--analysis-model`,
	trace analysis uses `--proposal-model` before falling back to the shared model.
	Without a shared model or `--judge-model`, each case keeps its configured judge.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[jharris1679]

@greptileai review