feat(ai-research-workflows): make iterate-plan a floating utility

[cdcore09]

Summary

Refactor iterate-plan from a fixed pipeline position to an explicitly phase-agnostic utility callable after any phase that reveals plan deficiencies. Update all commands to reference iterate-plan as available.

Requirements

• Add # When to Use iterate-plan section listing triggers from every phase (research, experiment, prototype, validate, optimize, requirement changes)
• Update iteration philosophy to emphasize: "iterate-plan is a tool, not a pipeline step"
• Add new iteration patterns: Incorporating Prototype Findings, Post-Validate Adjustment, Post-Optimize Escalation
• Update cross-reference section to list all phases that can trigger iterate-plan
• Update commands/research.md: add note about using iterate-plan when findings contradict existing plans
• Update commands/implement.md: mismatch handling should explicitly suggest /iterate-plan
• Verify experiment, prototype, validate, and optimize commands reference iterate-plan in their retreat paths (from Phases 2-4, 6)

Context

Phase 5 of the workflow redesign plan. Depends on Phases 2-4 (retreat paths in experiment, prototype, validate should be in place).

Currently iterate-plan is presented at a fixed position between experiment and implement in the workflow diagrams, despite being useful after any phase. This refactoring makes the floating nature explicit.

Implementation Notes

Key files:

• plugins/ai-research-workflows/commands/iterate-plan.md (primary changes)
• plugins/ai-research-workflows/commands/research.md (add reference)
• plugins/ai-research-workflows/commands/implement.md (update mismatch section, ~lines 69-101)