docs: add continuous learning roadmap and convert backlog to OpenSpec index

[tryweb]

Summary

• Add docs/roadmap.md with a 4-track roadmap for OpenCode memory learning capabilities
• Convert docs/backlog.md to an OpenSpec index format for tracking BL items against change IDs
• Revise user/team identity design to be single-user-first with conditional multi-user enablement
• Align retry/recovery and checkpoint/resume items as integration layer with OpenCode/oh-my-openagent (not re-implementing execution engines)

Changes

docs/roadmap.md

• Track 1: User Learning — single-user first, user/team as optional conditional
• Track 2: Agent Experience Learning — episodic task memory + success pattern extraction
• Track 3: Trust, Quality, Governance — citation, freshness, decay, conflict detection
• Track 4: Product UX & Observability — learning dashboard, KPIs, eval harness
• Integration boundary contract (Section 7/8): Clear responsibility split between OpenCode, oh-my-openagent, and this plugin
• Phase 0–4 with priorities and success metrics

docs/backlog.md

• Converted from full spec backlog to OpenSpec index
• Each BL entry now has: BL-ID, Title, Priority, Status, OpenSpec Change ID, Spec Path, Notes
• All items start with Status: planned, OpenSpec Change ID: TBD
• Maintenance rules: update change IDs when OpenSpec artifacts are created, prefer OpenSpec artifacts on conflicts

Key Design Decisions

1. Single-user first: userId/teamId are optional metadata, not P0 hard requirements
2. Integration over duplication: retry/recovery and checkpoint/resume focus on learning-layer evidence, not re-implementing OpenCode/OMO execution engines
3. Backlog as navigation: backlog.md is now an index; specs and tasks live in OpenSpec artifacts