Fix flaky telemetry stats regression tests

In #6907 we supposed to fix the telemetry stats regression test
flakiness but unfortutanely it was not enough.

What happens is when the regression test runner setup the database for
regression tests be executed we stop the scheduler and if for some
reason a system job is running then it will be killed and it produces
records on `bgw_job_stat` and `bgw_job_stat_history` with the crash
information that can potentially lead to flaky test output.

To prevent it we cleanup system job stats metadata table just after stop
the scheduler.

test/sql/utils/testsupport.sql

@@ -5,6 +5,10 @@
 
 SELECT _timescaledb_functions.stop_background_workers();
 
+-- Cleanup any system job stats that can lead to flaky test
+DELETE FROM _timescaledb_internal.bgw_job_stat_history WHERE job_id < 1000;
+DELETE FROM _timescaledb_internal.bgw_job_stat WHERE job_id < 1000;
+
 CREATE SCHEMA IF NOT EXISTS test;
 GRANT USAGE ON SCHEMA test TO PUBLIC;
 

tsl/test/expected/telemetry_stats.out

@@ -535,16 +535,6 @@ SELECT jsonb_pretty(get_telemetry_report() -> 'stats_by_job_type');
          "total_duration_failures": "@ 4 secs",+
          "max_consecutive_failures": 1         +
      },                                        +
-     "policy_job_stat_history_retention": {    +
-         "total_runs": 1,                      +
-         "total_crashes": 1,                   +
-         "total_duration": "@ 0",              +
-         "total_failures": 0,                  +
-         "total_successes": 0,                 +
-         "max_consecutive_crashes": 1,         +
-         "total_duration_failures": "@ 0",     +
-         "max_consecutive_failures": 0         +
-     },                                        +
      "policy_refresh_continuous_aggregate": {  +
          "total_runs": 2,                      +
          "total_crashes": 0,                   +