tests/e2e_finjector: add teardown method

Previously we did not heal network failures when exiting tests that use
this base class (namely, AvailabilityTests). As possible failure types
included netem failures, that meant that all subsequent tests in a test
run used a node with crippled network, leading to flakiness.

Fixes redpanda-data#5980

Author: ztlpn

tests/rptest/tests/e2e_finjector.py

@@ -89,3 +89,9 @@ def _failure_injector_loop(self):
             self.redpanda.logger.info(
                 f"waiting {delay} seconds before next failure")
             time.sleep(delay)
+
+    def teardown(self):
+        self.enable_failures = False
+        if self.finjector_thread:
+            self.finjector_thread.join()
+        FailureInjector(self.redpanda)._heal_all()