Remove `mellea.backends.vllm`

[nrfulton]

• Remove mellea/backends/vllm.py.
• Remove vllm group from pyproject.toml
• Update tests as needed so that we are still testing mellea/backends/openai.py against a vllm endpoint, but do not need vllm in pyproject.

[psschwei]

Given that vllm doesn't work on Mac (yet?), this probably makes sense.