feat(code-review): add --model flag to override agent model selection

[kpatel513]

Summary

The plugin uses different models per step (Haiku for lightweight tasks, Sonnet for compliance, Opus for bug detection). Users with different cost/quality preferences have no way to override this. This adds a --model flag that overrides all agents to use the specified model.

Usage

# Use Sonnet for all agents (cost-conscious, still high quality)
/code-review --model sonnet

# Use Opus for all agents (maximum thoroughness)
/code-review --model opus --comment

# Default behavior unchanged (per-step model selection)
/code-review

Changes

plugins/code-review/commands/code-review.md

• Add --model argument handling before step 1: if provided, overrides all per-step model defaults; if invalid, stops with a clear error

plugins/code-review/README.md

• Document --model flag in Options section with valid values and default behavior

Design decisions

• All-or-nothing override: --model applies to every agent uniformly. Per-agent overrides (e.g. --bug-model) would add complexity with limited benefit.
• Invalid value → hard stop: Silently falling back to defaults on a typo would be confusing. Explicit error is better.
• All three values allowed: Users choosing haiku for everything accept the quality tradeoff. Not our call to block it.

Test plan

• /code-review --model sonnet — all agents use Sonnet
• /code-review --model opus — all agents use Opus
• /code-review --model invalid — stops with "unknown model — valid values are haiku, sonnet, opus"
• /code-review (no flag) — per-step model selection unchanged