thermalctld: Ignore exception when deleting chassisdb entry fails

After sonic-mgmt tests, it does a config consistency check and if the
config consistency check fails, it runs recover_chassis(). That
function simultaneously runs config_reload on the supervisor and all
linecards (in parallel). This means that the supervisor networking can
be in the process of restarting when thermalctld exits and goes
through the cleanup process of deleting entries from chassisdb.

sonic-thermalctld/scripts/thermalctld

@@ -559,8 +559,14 @@ class TemperatureUpdater(logger.Logger):
             table_keys = self.table.getKeys()
             for tk in table_keys:
                 self.table._del(tk)
-                if self.is_chassis_upd_required and self.chassis_table is not None:
-                    self.chassis_table._del(tk)
+                try:
+                    if self.is_chassis_upd_required and self.chassis_table is not None:
+                        self.chassis_table._del(tk)
+                except Exception as e:
+                    # On a chassis system it is possible we may lose connection
+                    # to the supervisor and chassisdb. If this happens then we
+                    # should simply remove our handle to chassisdb.
+                    self.chassis_table = None
         if self.phy_entity_table:
             phy_entity_keys = self.phy_entity_table.getKeys()
             for pek in phy_entity_keys: