feat: add QRSPI-inspired workflow with 5 alignment phases + execution

[mrsimpson]

Summary

• New QRSPI workflow: Adds qrspi.yaml — a QRSPI (QRISPY)-inspired workflow adapted for our YAML-based state machine architecture
• 6 phases: Questions → Research → Design Discussion → Structure Outline → Plan → Implement
• 5 alignment phases + 1 execution: Enforces documentation-only work during alignment (*.md, *.txt, *.adoc), full access during implementation
• Anti-rush mechanisms: Feature ticket hidden during Research phase, ~200-line brain dump enforced in Design Discussion, "C Header File" style structure outline

Key Design Decisions

Decision	Rationale
6 phases (not 8)	Work Tree removed — task organization integrated into Plan phase
No PR state	PR is a manual review process, not an agent phase
abandon_feature → questions	Clean reset point for all abandon transitions
implement_complete → questions	Iterative: returns to questions for next feature
Domain: code, Complexity: high	Reflects the 6-phase upfront investment

Changes

• resources/workflows/qrspi.yaml — New workflow definition (6 states, 15+ transitions)
• packages/core/test/unit/workflow-validation.test.ts — Added 'qrspi' to expected core workflows
• packages/visualizer/src/services/workflow-list.ts — Auto-generated by build (25 workflows now)

Verification

• All 735 tests pass (385 core + 286 server + 64 opencode) — zero regressions
• Workflow validates against state-machine-schema.json
• All states reachable from initial state (BFS analysis)
• File restrictions consistent across alignment vs execution phases