feat(context): enhance timeout handling, tenant isolation, and budget management

- Added timeout enforcement for token counting, scoring, and compression with detailed error handling.
- Introduced tenant isolation in context caching using project and agent identifiers.
- Enhanced budget management with stricter checks for critical context overspending and buffer limitations.
- Optimized per-context locking with cleanup to prevent memory leaks in concurrent environments.
- Updated default assembly timeout settings for improved performance and reliability.
- Improved XML escaping in Claude adapter for safety against injection attacks.
- Standardized token estimation using model-specific ratios.

backend/app/services/context/cache/context_cache.py

@@ -95,19 +95,28 @@ class ContextCache:
         contexts: list[BaseContext],
         query: str,
         model: str,
+        project_id: str | None = None,
+        agent_id: str | None = None,
     ) -> str:
         """
         Compute a fingerprint for a context assembly request.
 
         The fingerprint is based on:
+        - Project and agent IDs (for tenant isolation)
         - Context content hash and metadata (not full content for performance)
         - Query string
         - Target model
 
+        SECURITY: project_id and agent_id MUST be included to prevent
+        cross-tenant cache pollution. Without these, one tenant could
+        receive cached contexts from another tenant with the same query.
+
         Args:
             contexts: List of contexts
             query: Query string
             model: Model name
+            project_id: Project ID for tenant isolation
+            agent_id: Agent ID for tenant isolation
 
         Returns:
             32-character hex fingerprint
@@ -128,6 +137,9 @@ class ContextCache:
             )
 
         data = {
+            # CRITICAL: Include tenant identifiers for cache isolation
+            "project_id": project_id or "",
+            "agent_id": agent_id or "",
             "contexts": context_data,
             "query": query,
             "model": model,