explicitly set workers_use_ray to ensure vllm doesnt fallback to mp backend, leading to interference with communication with evaluators

- /models/dummy_model.py
- /models/rag_llama_baseline.py
- /models/vanilla_llama_baseline.py

Batch predict interface v0

See merge request aicrowd/challenges/meta-comprehensive-rag-benchmark-kdd-cup-2024/meta-comphrehensive-rag-benchmark-starter-kit!4

add baseline related docs

See merge request aicrowd/challenges/meta-comprehensive-rag-benchmark-kdd-cup-2024/meta-comphrehensive-rag-benchmark-starter-kit!1