refactor: transaction() as an asynccontextmanager over asyncpg's own contexts

T3-21: the _Transaction class hand-rolled acquire / start / commit-or-rollback /
release plus a leak guard for a failed start. an asynccontextmanager over the
driver's own pool.acquire() context does the same, and the acquire context
releases on every path including a failed start. verified on a real backend:
commit, rollback + re-raise, cancellation rollback, failed-start leak guard
(pool max 1) and release after 50 transactions.

Signed-off-by: disqualifier <dev@disqualifier.me>

src/psql/psql.py

@@ -44,6 +44,7 @@ errors (FAIL LOUD - unlike the mongo lib's swallow-and-default):
 
 import asyncio
 import logging
+from contextlib import asynccontextmanager
 from typing import Any, Dict, List, Optional, Sequence
 
 import asyncpg
@@ -236,7 +237,7 @@ class PsqlDB:
         commits on clean exit, rolls back and re-raises on any error. `conn` is a raw
         asyncpg connection ($1-placeholder execute/fetch/... directly).
         """
-        return _Transaction(self.pool)
+        return _transaction(self.pool)
 
     # -------------------------------------------------------------------------
     #   layer 1 - friendly portable verbs (identical across psql/mysql)
@@ -365,36 +366,17 @@ class PsqlDB:
         return _status_count(status)
 
 
-class _Transaction:
-    """async ctx mgr: acquire a pooled connection + open a transaction on it"""
+@asynccontextmanager
+async def _transaction(pool: asyncpg.Pool):
+    """acquire a pooled connection and run the block in a transaction on it
 
-    def __init__(self, pool: asyncpg.Pool):
-        self._pool = pool
-        self._conn = None
-        self._tx = None
-
-    async def __aenter__(self):
-        self._conn = await self._pool.acquire()
-        try:
-            self._tx = self._conn.transaction()
-            await self._tx.start()
-        except BaseException:
-            # start() (or transaction()) failing after acquire would otherwise leak the
-            # pooled connection - __aexit__ is not called when __aenter__ raises. release
-            # it and reset so a failed transaction start never burns a pool slot.
-            await self._pool.release(self._conn)
-            self._conn = None
-            raise
-        return self._conn
-
-    async def __aexit__(self, exc_type, exc, tb) -> None:
-        try:
-            if exc_type is None:
-                await self._tx.commit()
-            else:
-                await self._tx.rollback()
-        finally:
-            await self._pool.release(self._conn)
+    asyncpg's own context managers do the work: the transaction commits on clean exit
+    and rolls back on any error, and the acquire releases the connection on every path -
+    including a failed transaction start, so a failed start never burns a pool slot
+    """
+    async with pool.acquire() as conn:
+        async with conn.transaction():
+            yield conn
 
 
 def _where(conditions: Optional[Dict[str, Any]]) -> tuple: