Meta: Make GA logs less misleading

If the job fails early (e.g. during linting), the 'cat debug.log' step would *also* fail.
This would confuse GA into thinking that this is the crucial thing and highlights it.
This misleads the user into looking at the wrong thing.

.github/workflows/cmake.yml

@@ -143,14 +143,22 @@ jobs:
       env:
         SERENITY_KERNEL_CMDLINE: "boot_mode=self-test"
         SERENITY_RUN: "ci"
-      run: ninja run
+      run: |
+        echo "::group::ninja run # Qemu output"
+        run_worked=y
+        ninja run || run_worked=n
+        echo "::endgroup::"
+        echo "::group::cat debug.log # Serenity output"
+        if [ "y" = "${run_worked}" ] ; then
+          cat debug.log
+        else
+          echo "skipped (qemu had non-zero exit-code)"
+        fi
+        echo "::endgroup::"
+        [ "y" = ${run_worked} ]
       timeout-minutes: 10
       # FIXME: When stable, remove continue on error. (See issue #5541)
       continue-on-error: true
-    - name: Print target logs
-      if: ${{ !cancelled() && matrix.debug-macros == 'NORMAL_DEBUG'}}
-      working-directory: ${{ github.workspace }}/Build
-      run: cat ./debug.log
 
   build_and_test_lagom:
     runs-on: ${{ matrix.os }}