fix: propagate EmbeddingModelMismatchError through MCP tools

@NickShtefan fixed the CLI path in PR MemPalace#442 (searcher.py) by re-raising
EmbeddingModelMismatchError past the generic except clause. The MCP
path had the same bug via two additional swallows:

- mcp_server.py:_get_collection() wraps _palace_get_collection in
  except Exception: return None, so every tool using _get_collection
  (status, list_wings, list_rooms, get_taxonomy, check_duplicate,
  search, etc.) silently showed "No palace found" instead of the
  actual mismatch error.

- The top-level dispatcher in handle_request() catches Exception and
  returns hardcoded "Internal tool error", so even if the wrapper
  re-raises, the actionable message (stored vs current model + recovery
  instructions) never reaches the caller.

Fix in three spots:
1. searcher.py: re-raise EmbeddingModelMismatchError in search() and
   search_memories() before the generic except (matches @NickShtefan's fix).
2. mcp_server.py:_get_collection(): re-raise EmbeddingModelMismatchError
   before the generic swallow.
3. mcp_server.py handle_request(): catch EmbeddingModelMismatchError
   explicitly and return the message as a JSON result so MCP clients
   see the real error and recovery instructions.

Author: FabioLissi

mempalace/mcp_server.py

@@ -26,7 +26,12 @@
 from datetime import datetime
 from pathlib import Path
 
-from .config import MempalaceConfig, sanitize_name, sanitize_content
+from .config import (
+    EmbeddingModelMismatchError,
+    MempalaceConfig,
+    sanitize_name,
+    sanitize_content,
+)
 from .version import __version__
 from .palace import get_collection as _get_collection_from_palace
 from .query_sanitizer import sanitize_query
@@ -111,6 +116,8 @@ def _get_collection(create=False):
                 create=create,
             )
         return _collection_cache
+    except EmbeddingModelMismatchError:
+        raise
     except Exception:
         return None
 
@@ -981,6 +988,17 @@ def handle_request(request):
                 "id": req_id,
                 "result": {"content": [{"type": "text", "text": json.dumps(result, indent=2)}]},
             }
+        except EmbeddingModelMismatchError as e:
+            logger.error(f"Embedding model mismatch in {tool_name}: {e}")
+            return {
+                "jsonrpc": "2.0",
+                "id": req_id,
+                "result": {
+                    "content": [
+                        {"type": "text", "text": json.dumps({"error": str(e)}, indent=2)}
+                    ]
+                },
+            }
         except Exception:
             logger.exception(f"Tool error in {tool_name}")
             return {