Split generic IO Processor plugins tests from Terratorch specific ones (#35756)

Signed-off-by: Christian Pinto <christian.pinto@ibm.com>

tests/conftest.py

@@ -1140,6 +1140,15 @@ class VllmRunner:
         return self
 
     def __exit__(self, exc_type, exc_value, traceback):
+        # Explicitly shutdown the engine core to release GPU resources
+        # This is needed because when executing consecutive tests, the GC
+        # might not be fast enough in shutting down the llm engine. This can lead to OOMs
+        # because when the next test starts some GPU memory is still in use.
+        try:
+            self.llm.llm_engine.engine_core.shutdown()
+        except Exception:
+            # Ignore shutdown errors as cleanup will still proceed
+            pass
         del self.llm
         cleanup_dist_env_and_memory()