Fix bgw_custom flakiness by stopping bgws

When dropping a table, we get notices if that causes to cancel
background jobs and thus fail tests. By stopping the workers
before dropping the table, we should reduce this flakiness.

(cherry picked from commit 6e4d174)

tsl/test/expected/bgw_custom.out

@@ -972,7 +972,19 @@ SELECT count(*) = 0
 (1 row)
 
 -- cleanup
+SELECT _timescaledb_functions.stop_background_workers();
+ stop_background_workers 
+-------------------------
+ t
+(1 row)
+
 DROP TABLE sensor_data;
+SELECT _timescaledb_functions.restart_background_workers();
+ restart_background_workers 
+----------------------------
+ t
+(1 row)
+
 -- Github issue #5537
 -- Proc that waits until the given job enters the expected state
 CREATE OR REPLACE PROCEDURE wait_for_job_status(job_param_id INTEGER, expected_status TEXT, spins INTEGER=:TEST_SPINWAIT_ITERS)

tsl/test/sql/bgw_custom.sql

@@ -629,7 +629,9 @@ SELECT count(*) = 0
   WHERE hypertable_name = 'sensor_data' AND NOT is_compressed;
 
 -- cleanup
+SELECT _timescaledb_functions.stop_background_workers();
 DROP TABLE sensor_data;
+SELECT _timescaledb_functions.restart_background_workers();
 
 -- Github issue #5537
 -- Proc that waits until the given job enters the expected state