fix: implement pending_messages cleanup to prevent unbounded growth

Fixes #353 - Observations not being saved due to incomplete pending messages implementation

Changes:
- PendingMessageStore.markProcessed() now clears tool_input and tool_response
- PendingMessageStore.cleanupProcessed() changed from time-based to count-based retention
- Keeps most recent 100 processed messages for UI display
- SDKAgent.processSDKResponse() calls cleanup after marking messages processed

This prevents the database from growing unbounded with duplicate transcript data.
The pending_messages table now only stores full transcripts for pending/processing
messages, while processed messages keep metadata only.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

CLAUDE.md

@@ -76,3 +76,7 @@ Settings are managed in `~/.claude-mem/settings.json`. The file is auto-created
 **Public Docs**: https://docs.claude-mem.ai (Mintlify)
 **Source**: `docs/public/` - MDX files, edit `docs.json` for navigation
 **Deploy**: Auto-deploys from GitHub on push to main
+
+# Important
+
+No need to edit the changelog ever, it's generated automatically.