[feature]Implement session cache for Doris connections (#44)

* [feature]Implement session cache for Doris connections

This PR introduces a `DorisSessionCache` to cache and reuse `DorisConnection` objects in memory.
This helps to reduce the overhead of creating new connections, especially for frequently used system sessions like "query"
and "system", and avoid not calling release_connection leads to `Connection acquisition timed out` when the number of
connection pools reaches the maximum value.

The PR #34 fixed the issue when calling the tool `exec_query`, but in the codebase, a large number of other tools directly
using get_connection ("query") to get connection object but without calling the release_connection method will cause the
connection to fail to be obtained after a certain number of times.

Key changes:
- Added `DorisSessionCache` class to manage the lifecycle of cached sessions.
- The cache is configurable to store system sessions, user sessions, or both. By default, only system sessions are cached.
- Integrated the session cache into `DorisConnectionManager`.
- `get_connection` now checks the cache before creating a new connection.
- `release_connection` removes the connection from the cache.

* Add tests

doris_mcp_server/utils/analysis_tools.py

@@ -516,7 +516,7 @@ class SQLAnalyzer:
             try:
                 # Switch to specified database/catalog if provided
                 if catalog_name:
-                    await connection.execute(f"USE `{catalog_name}`")
+                    await connection.execute(f"SWITCH `{catalog_name}`")
                 if db_name:
                     await connection.execute(f"USE `{db_name}`")
                 
@@ -1237,4 +1237,4 @@ class MemoryTracker:
                 "tracker_names": tracker_names,
                 "time_range": time_range,
                 "timestamp": datetime.now().isoformat()
-            } 
+            }