Fix SearchIndexService OOM — rebuild guard + memory cap#

Problem#

SearchIndexService.buildIndex() had no concurrency guard. Rapid file changes triggered overlapping full rebuilds, each allocating a ~24K-document index Map, causing OOM crashes.

Task / Link#

GitHub Issue #14 — AgentSDE/meridian-backend

Changes#

• Add AbortController-based rebuild guard: concurrent handleFileChange() calls abort the in-flight rebuild and set rebuildPending instead of spawning a second full build
• After a rebuild completes or is aborted, check rebuildPending and trigger one follow-up rebuild if needed
• Skip handleFileChangeDetail() incremental updates when rebuildPending is true (upcoming full rebuild covers them)
• Add node_args: '--max-old-space-size=512' and max_memory_restart: '512M' to ecosystem.config.js
• Add --max-old-space-size=512 to exec node line in services/start-backend.sh
• Add unit tests: concurrent guard, abort-preserves-index, follow-up rebuild, incremental-skip-when-pending

Notes#

• The 500ms debounce on FILE_CHANGE_EVENT in FileWatcherService is unchanged and still applies
• Single rebuild peak memory is well under 512MB — OOM was caused by concurrent allocations, not a single run
• signal.aborted is checked per-task (not per-file) to avoid per-file overhead with small markdown files

Testing#

• All 655 unit tests pass (npm run test)
• Lint clean — zero errors (npm run lint)