fuzz: force fuzzing job on CI to fail if it founds any errors (#7756)

Pro: you can easily detect if you have errors in your code base.

Cons: the CI might fail for unrelated errors, i.e. for errors unrelated
to the code you are trying to push/commit/merge. And that *will* happen a
lot, especially just after this PR has been merged, because there are
lots of errors right now.

Memory leak detection is still disabled.

This is the behavior that we have on nDPI too.

.github/workflows/cifuzz.yml

@@ -9,20 +9,26 @@ jobs:
         sanitizer: [address]
     steps:
     - name: Build Fuzzers - ${{ matrix.sanitizer }}
-      id: build
       uses: google/oss-fuzz/infra/cifuzz/actions/build_fuzzers@master
       with:
         oss-fuzz-project-name: 'ntopng'
+        dry-run: false
         sanitizer: ${{ matrix.sanitizer }}
     - name: Run Fuzzers - ${{ matrix.sanitizer }}
       uses: google/oss-fuzz/infra/cifuzz/actions/run_fuzzers@master
       with:
         oss-fuzz-project-name: 'ntopng'
+        fuzz-seconds: 2200
+        dry-run: false
         sanitizer: ${{ matrix.sanitizer }}
-        fuzz-seconds: 1200
+    - name: Check Crash (fails when a crash is detected)
+      # Run Fuzzers return success even when setting dry-run to false.
+      # A temporal workaround is to trigger failure manually if we fing crash files.
+      run: |
+        exit $(ls out/artifacts |wc -l)
     - name: Upload Crash
-      uses: actions/upload-artifact@v3
-      if: failure() && steps.build.outcome == 'success'
+      uses: actions/upload-artifact@v1
+      if: failure()
       with:
         name: ${{ matrix.sanitizer }}-artifacts
-        path: ./out/artifacts
+        path: ./out/artifacts