[llm][test] Use smolVLM in unit tests

[lk-chen]

Why are these changes needed?

While bumping vLLM to 0.8.2, test_vision_model is failing because:

• the previously used llava-354m model was random generated
• during test, multi_modal_projector(..) returns [[nan, nan, ...]...]
• this line introduced in vLLM 0.8.* does not like nan, resulting an empty mm_embedding, and further failures
• so instead of using fake model (llava-354m), let's start using a real yet small vision model (smolVLM)

The model is copied from HuggingFaceTB/SmolVLM-256M-Instruct following this instruction

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(
  • Manually tested: install vllm0.8.2 locally, unit tests pass

[GeneDer]