feat: Redis opportunity cache, CW API retry/logging, adaptive TTLs

- Add Redis-backed opportunity cache with background refresh (30s interval)
- Fix concurrency bug: use lazy thunks instead of eager promises for batching
- Add withCwRetry utility with exponential backoff for transient CW errors
- Add adaptive TTL algorithms (primary, sub-resource, products) based on opportunity activity
- Add include query param on GET /sales/opportunities/:id (notes,contacts,products)
- Add opt-in CW API logger (LOG_CW_API env var) with timestamped files in cw-api-logs/
- Add debug-scripts/analyze-cw-calls.py for API call analysis
- Add computeSubResourceCacheTTL and computeProductsCacheTTL algorithms with tests
- Increase CW API timeout from 15s to 30s
- Unblock cache refresh from startup chain (remove await)
- Prioritize recently updated opportunities in refresh cycle
- Add CACHING.md documentation
- Update API_ROUTES.md with caching details and include param
- Update copilot instructions to require CACHING.md sync
- Add dev:log script for CW API call logging during development

src/modules/globalEvents.ts

@@ -186,6 +186,10 @@ interface EventTypes {
     totalOpportunities: number;
     activitiesRefreshed: number;
     companiesRefreshed: number;
+    notesRefreshed: number;
+    contactsRefreshed: number;
+    productsRefreshed: number;
+    oppCwDataRefreshed: number;
     skipped: number;
   }) => void;
   "cache:opportunities:refresh:error": (data: { error: unknown }) => void;