feat: integrate Spark Intelligence into Timmy swarm system

Adds a self-evolving cognitive layer inspired by vibeship-spark-intelligence,
adapted for Timmy's agent architecture. Spark captures swarm events, runs
EIDOS prediction-evaluation loops, consolidates memories, and generates
advisory recommendations — all backed by SQLite consistent with existing
patterns.

New modules:
- spark/memory.py — event capture with importance scoring + memory consolidation
- spark/eidos.py — EIDOS cognitive loop (predict → observe → evaluate → learn)
- spark/advisor.py — ranked advisory generation from accumulated intelligence
- spark/engine.py — top-level API wiring all subsystems together

Dashboard:
- /spark/ui — full Spark Intelligence dashboard (3-column: status/advisories,
  predictions/memories, event timeline) with HTMX auto-refresh
- /spark — JSON API for programmatic access
- SPARK link added to navigation header

Integration:
- Coordinator hooks emit Spark events on task post, bid, assign, complete, fail
- EIDOS predictions generated when tasks are posted, evaluated on completion
- Memory consolidation triggers when agents accumulate enough outcomes
- SPARK_ENABLED config toggle (default: true)

Tests: 47 new tests covering all Spark subsystems + dashboard routes.
Full suite: 538 tests passing.

https://claude.ai/code/session_01KJm6jQkNi3aA3yoQJn636c

pyproject.toml

@@ -72,6 +72,7 @@ include = [
     "src/notifications",
     "src/shortcuts",
     "src/telegram_bot",
+    "src/spark",
 ]
 
 [tool.pytest.ini_options]