Update e2e/lm-eval test infrastructure

[dbarbuzzi]

SUMMARY:
Remove some no longer necessary flags from the run_tests.sh script and update e2e/lm-eval test infra to use pytest’s parametrization.

Unused flags

A couple of the flags in the test script were added to support reporting efforts in our CI/CD, but they are no longer necessary as they are handled outside the test script.

e2e/lm-eval test parametrization

This change is primarily to improve the reporting for these sets of tests, particularly about the test name that is generated. There are some challenges that the current, non-pytest parametrization approach introduce:

1. The parametrization index is included in the generated name. This is problematic because any change to the order of the parametrization then results in breaking references (e.g., historic views of the test results) between past and future runs.
   • Technically, this is not an issue in these two examples because each config is executed individually, but it would be great to transition any other instances to pytest’s built-in parametrization to help with our test tracking.
2. Parametrizing a class results in a 'proper' class name, which normalizes certain values. Specifically, the parameters here are file paths which include slashes which are in turn normalized, replacing the slashes with different characters.
   • Alternatively, pytest puts all parametrization in the test name itself and doesn't do any of this normalization, preserving the original data.

For point 2, an example of the before/after naming:

Before
classname: tests/e2e/vLLM/test_vllm.py::TestvLLM_0_tests_e2e_vLLM_configs_kv_cache_tinyllama_yaml
test name: test_vllm
full name: tests/e2e/vLLM/test_vllm.py::TestvLLM_0_tests_e2e_vLLM_configs_kv_cache_tinyllama_yaml::test_vllm

After
classname: tests/e2e/vLLM/test_vllm.py::TestvLLM
test name: test_vllm[tests/e2e/vLLM/configs/kv_cache_tinyllama.yaml]
full name: tests/e2e/vLLM/test_vllm.py::TestvLLM::test_vllm[tests/e2e/vLLM/configs/kv_cache_tinyllama.yaml]

TEST PLAN:
There are two (internal) test runs, one e2e and one lm-eval, showing the new changes working without issue to be communicated internally.

[github-actions]

👋 Hi! Thank you for contributing to llm-compressor. Please add the ready label when the PR is ready for review.

Note: This is required to complete the testing suite, please only add the label once the PR is code complete and local testing has been performed.