divide out tensor engine tests (#649)

Summary:
Pull Request resolved: #649

Rollback Plan:

Differential Revision: D78989844

Author: allenwang28

.github/workflows/test-cpu.yml

@@ -34,7 +34,4 @@ jobs:
         # Install the built wheel from artifact
         install_wheel_from_artifact
 
-        # Currently a no-op.
-        # Tests requiring tensor engine / GPU need to be identified and flagged to skip.
-        # We will just ensure monarch can be imported successfully.
-        python -c "import monarch; print('Monarch imported successfully')"
+        LC_ALL=C pytest python/tests/ -s -v -m "not oss_skip" --ignore=python/tests/tensor_engine -n 4

.github/workflows/test-cuda.yml

@@ -55,5 +55,5 @@ jobs:
         pyright python/tests/test_python_actors.py
 
         # Run CUDA tests
-        LC_ALL=C pytest python/tests/ -s -v -m "not oss_skip"
+        LC_ALL=C pytest python/tests/tensor_engine -s -v -m "not oss_skip" -n 4
         python python/tests/test_mock_cuda.py

python/tests/test_coalescing.py renamed to python/tests/tensor_engine/test_coalescing.py

@@ -78,7 +78,7 @@ def local_device_mesh(
             num_hosts,
             gpu_per_host,
             activate,
-            backend=str(backend_type),
+            backend=backend_type.value,
         )
 
     @property