feat: use connection pool instead of a single shared connection

This is useful for multitask environments (like the API), to avoid
conflicts

server/app/adapters/postgres_query.py

@@ -2,12 +2,12 @@ from datetime import datetime
 from dateutil.relativedelta import relativedelta
 
 import psycopg2
-from psycopg2.extensions import connection
 
 from domain.entities import ConsumptionPoint
 from domain.exceptions import DatabaseError
 from domain.value_objects import Granularity
 from ports.reading_query_repository import ReadingQueryRepository
+from infrastructure.db import get_conn
 
 
 _GRANULARITY_DELTA = {
@@ -19,9 +19,6 @@ _GRANULARITY_DELTA = {
 
 
 class PgReadingQueryRepository(ReadingQueryRepository):
-    def __init__(self, conn: connection) -> None:
-        self._conn = conn
-
     def get_consumption(
         self,
         dev_eui: str,
@@ -54,9 +51,10 @@ class PgReadingQueryRepository(ReadingQueryRepository):
             ORDER BY period ASC
         """
         try:
-            with self._conn.cursor() as cur:
-                cur.execute(query, (date_trunc, dev_eui, adjusted_start, end))
-                rows = cur.fetchall()
+            with get_conn() as conn:
+                with conn.cursor() as cur:
+                    cur.execute(query, (date_trunc, dev_eui, adjusted_start, end))
+                    rows = cur.fetchall()
         except psycopg2.DatabaseError as e:
             raise DatabaseError(f"Erreur requête consumption : {e}") from e