Enable faulthandler to dump stack on critical failure (#8207)

With luck, will dump on signals and help us get a backtrace.

edb/tools/test/runner.py

@@ -26,6 +26,7 @@
 import csv
 import dataclasses
 import enum
+import faulthandler
 import io
 import itertools
 import json
@@ -69,6 +70,8 @@
 py_hash_secret: bytes = cpython_state.get_py_hash_secret()
 py_random_seed: bytes = random.SystemRandom().randbytes(8)
 
+faulthandler.enable(file=sys.stderr, all_threads=True)
+
 
 def teardown_suite() -> None:
     # The TestSuite methods are mutating the *result* object,
@@ -92,6 +95,10 @@ def init_worker(
     global py_hash_secret
     global py_random_seed
 
+    faulthandler.enable(file=sys.stderr, all_threads=True)
+    # If we're still running after 30 minutes, dump the traceback.
+    faulthandler.dump_traceback_later(30 * 60, file=sys.stderr)
+
     if additional_init:
         additional_init()