Update on "Add preprocess to ci"

1. Run `test_preprocess.py` in ci.
2. Refactor `test_preprocess.py` by creating the model artifacts (eager, exported, et) once and re-use them on different inputs (previously, create model artifacts for each test).

Test Plan:
```
pytest -c /dev/null -v -n auto examples/models/llama3_2_vision/preprocess/
```

[ghstack-poisoned]

Author: lucylq

.github/workflows/_unittest.yml

@@ -38,7 +38,7 @@ jobs:
         .ci/scripts/setup-linux.sh cmake
 
         # Install llama3_2_vision dependencies.
-        bash examples/models/llama3_2_vision/install_requirements.sh
+        PYTHON_EXECUTABLE=python ./examples/models/llama3_2_vision/install_requirements.sh
 
         # Run pytest with coverage
         pytest -n auto --cov=./ --cov-report=xml
@@ -70,6 +70,8 @@ jobs:
         ${CONDA_RUN} --no-capture-output \
         .ci/scripts/setup-macos.sh cmake
 
+        ${PYTHON_EXECUTABLE} pip list
+
         # Install llama3_2_vision dependencies.
         bash examples/models/llama3_2_vision/install_requirements.sh
 

examples/models/llama3_2_vision/install_requirements.sh

@@ -8,6 +8,7 @@
 # Install torchtune nightly for model definitions.
 echo "lfq check pip"
 pip list # debugging
+echo "lfq end pip"
 pip install --pre torchtune --extra-index-url https://download.pytorch.org/whl/nightly/cpu --no-cache-dir
 
 # Install torchao.