Print gdb stack traces from core dumps on failures

Because:
- Although we upload core dump artifacts, it's hard to debug core dumps
  on another system than the one that generated them.

- Using the original system to just print the stack traces from any core
  dumps found can help a lot to figure out what's going on.

.github/workflows/build-and-test.yml

@@ -97,6 +97,19 @@ jobs:
           path: |
             /tmp/core-*
 
+      - name: Print stack traces of core dumps (${{ matrix.python-version }})
+        if: failure()
+        run: |
+          sudo apt-get install -y gdb
+          set -x
+          for core in /tmp/core-*; do
+            echo "===== CORE: $core ====="
+            file "$core"
+            program=$(gdb -ex quit -c "$core" | grep -o "generated by.*" | awk '{print $3}' | tr -d '`')
+            gdb -ex bt "$program" "$core"
+            echo
+          done
+
   publish-test-results:
     name: "Publish test results"
     needs: build-and-test