feat: Implement comprehensive OpenAI error handling for Issue #362

Replace silent failures with clear, actionable error messages to eliminate
90-minute debugging sessions when OpenAI API quota is exhausted.

## Backend Enhancements
- Add error sanitization preventing sensitive data exposure (API keys, URLs, tokens)
- Add upfront API key validation before expensive operations (crawl, upload, refresh)
- Implement fail-fast pattern in RAG service (no more empty results for API failures)
- Add specific error handling for quota, rate limit, auth, and API errors
- Add EmbeddingAuthenticationError exception with masked key prefix support

## Frontend Enhancements
- Create enhanced error utilities with OpenAI-specific parsing
- Build TanStack Query compatible API wrapper preserving ETag caching
- Update knowledge service to use enhanced error handling
- Enhance TanStack Query hooks with user-friendly error messages

## Security Features
- Comprehensive regex sanitization (8 patterns) with ReDoS protection
- Input validation and circular reference detection
- Generic fallback messages for sensitive keywords
- Bounded quantifiers to prevent regex DoS attacks

## User Experience
- Clear error messages: "OpenAI API quota exhausted"
- Actionable guidance: "Check your OpenAI billing dashboard and add credits"
- Immediate error visibility (no more silent failures)
- Appropriate error severity styling

## Architecture Compatibility
- Full TanStack Query integration maintained
- ETag caching and optimistic updates preserved
- No performance regression (all existing tests pass)
- Compatible with existing knowledge base architecture

Resolves #362: Users no longer experience mysterious empty RAG results
that require extensive debugging to identify OpenAI quota issues.

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

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

python/src/agents/base_agent.py

@@ -216,7 +216,7 @@ class BaseAgent(ABC, Generic[DepsT, OutputT]):
             self.logger.info(f"Agent {self.name} completed successfully")
             # PydanticAI returns a RunResult with data attribute
             return result.data
-        except asyncio.TimeoutError:
+        except TimeoutError:
             self.logger.error(f"Agent {self.name} timed out after 120 seconds")
             raise Exception(f"Agent {self.name} operation timed out - taking too long to respond")
         except Exception as e: