Show compaction summary after context compression

Problem

During long implementation sessions (multi-hour feature builds, plan execution with code review cycles), context compaction happens silently. After compaction, there's no indication of what was discussed or decided in the compressed portion. This makes it hard to orient after compaction — especially when decisions, NLM consultations, or architectural choices were made earlier in the session.

Proposed Solution

After compaction completes, display a brief summary banner in the conversation, e.g.:

★ Context Compacted ─────────────────────────────
- Implemented DISC module (seae.py, disc.py) across 8 tasks
- Ran 50-card test (24% DISC trigger rate, 0 verdict changes)
- 3 NLM consultations: SEAE threshold, accept-if-better tie, CONTRADICTED trigger
- Code review found critical SEAE gate bug → fixed
- Commits: fed3eb2, 03597fa, f67a4fa, a1f148c, c6cad6f, e46f980
─────────────────────────────────────────────────

This would be generated by the model as part of the compaction process — no hook needed. Just include a visible summary block in the output after the compressed context is applied.

Why It Matters

• Long sessions accumulate decisions that inform later work. Losing visibility into those decisions mid-session causes repeated questions and inconsistent behavior.
• Users doing plan-driven development (brainstorm → plan → execute → review → merge) regularly hit compaction during the execute phase. The summary would preserve continuity.
• It's a quality-of-life improvement that costs almost nothing (the model already has the context it's compacting — summarizing it is trivial).

Alternatives Considered

• PostCompaction hook event: More flexible but requires user setup. A built-in summary is simpler and benefits everyone.
• Manual /compact command with summary: Only helps when compaction is user-initiated, not when it happens automatically.