[tests] move recipe parsing test so it is run nightly (#1825)

SUMMARY:
`text_recipe_parsing.py` was previously in `tests/llmcompressor/recipe`,
but since it is decorated with `@requires_gpu` it was never being run.
`llm-compressor-testing` only runs these tests if marked as
`integration` tests and they are inside
`tests/llmcompressor/transformers/(compression|finetune|obcq)` (see
[here](https://github.com/neuralmagic/llm-compressor-testing/blob/main/.github/actions/test/action.yml#L28)).

This moves the test into compression and marks it as an integration test
so it will be included in tests. This is valid since it is calling into
oneshot and running a compression flow on the model.


TEST PLAN:
- [x] Confirm with @dbarbuzzi that these tests are now being run in
nightly. Tests should succeed now that #1810 is merged in

---------

Signed-off-by: Brian Dellabetta <[email protected]>

Author: brian-dellabetta

tests/llmcompressor/recipe/recipe.yaml renamed to tests/llmcompressor/transformers/compression/recipes/smoothquant_gptq_w8a8.yaml

@@ -67,13 +67,13 @@ def recipe_variants():
       scheme: W8A8
 """
 
-    yaml_path = str(Path(__file__).parent / "recipe.yaml")
+    yaml_path = str(Path(__file__).parent / "recipes" / "smoothquant_gptq_w8a8.yaml")
 
     return [python_modifiers, yaml_string, yaml_path]
 
 
 @requires_gpu
-@pytest.mark.regression
+@pytest.mark.integration
 @pytest.mark.parametrize(
     "recipe",
     recipe_variants(),