Try multiproc in test

To fix the test I switched to uniproc processor, but
now I'm getting weird issues like

```
torch._dynamo.exc.BackendCompilerFailed: backend='<vllm.compilation.backends.VllmBackend object at 0x7ef9f68ca600>' raised:
AttributeError: module 'torch._tensor' has no attribute 'split'
```

Signed-off-by: Max de Bayser <mbayser@br.ibm.com>

Author: maxdebayser

tests/model_executor/test_model_load_with_params.py

@@ -26,7 +26,8 @@ def test_model_loading_with_params(vllm_runner, monkeypatch):
     """
     Test parameter weight loading with tp>1.
     """
-    monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
+    # to use apply_model
+    monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
     with vllm_runner(model_name=MODEL_NAME,
                      revision=REVISION,
                      dtype="float16",
@@ -66,7 +67,8 @@ def test_roberta_model_loading_with_params(vllm_runner, monkeypatch):
     """
     Test parameter weight loading with tp>1.
     """
-    monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
+    # to use apply_model
+    monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
     with vllm_runner(model_name=MODEL_NAME_ROBERTA,
                      revision=REVISION_ROBERTA,
                      dtype="float16",
@@ -108,7 +110,7 @@ def test_facebook_roberta_model_loading_with_params(vllm_runner, monkeypatch):
     Test loading roberta-base model with no lm_head.
     """
     # to use apply_model
-    monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
+    monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
     model_name = "FacebookAI/roberta-base"
     with vllm_runner(model_name=model_name,
                      dtype="float16",