refactor: Codify parallel execution rules in .ai-rules

[JeremyDev87]

Purpose

Move parallel execution learnings from personal memory files to shared .ai-rules so all AI tools (Cursor, Codex, Q, Kiro) benefit from the same rules.

Changes

• packages/rules/.ai-rules/rules/parallel-execution.md — New rule file with parallel execution guidelines
• packages/rules/.ai-rules/rules/core.md — Add reference to parallel-execution rules

Acceptance Criteria

• File overlap detection rule: "Issues modifying the same files MUST NOT run in the same parallel wave"
• Wave planning guidelines: dependency analysis → grouping → ordering
• Post-execution verification checklist
• Rules are tool-agnostic (not Claude Code specific)
• Existing tests pass
• Can be merged independently

[JeremyDev87]

Technical Details

Exact Changes

Create packages/rules/.ai-rules/rules/parallel-execution.md:

# Parallel Execution Rules

## Iron Law: Zero-Conflict Guarantee
Issues modifying the same files MUST NOT run in the same parallel wave.

## Wave Planning
1. Extract target files from each issue
2. Build conflict graph (shared files = edge)
3. Group non-conflicting issues into waves
4. Order waves by dependency chain

## File Overlap Detection
Before dispatching parallel agents:
- [ ] List all files each issue will modify
- [ ] Check for intersections between issue file sets
- [ ] If overlap found: separate into different waves or merge issues

## Post-Execution Verification
- [ ] All PRs pass CI independently
- [ ] No merge conflicts between PRs in the same wave
- [ ] Context document updated with results

Steps

1. Create the rule file
2. Add reference in core.md under a "Related Rules" section
3. Verify rules are discoverable via search_rules MCP tool
4. Remove redundant content from personal memory (feedback_no_file_overlap_parallel.md) after rule is live

Background

Learned from 3-Wave execution of 12 issues: PR #750 and #751 conflicted because they modified the same agent JSON files. Had to close both and re-execute sequentially. This rule prevents that waste.