Add vLLM version info to logs and openai API server

[jasonacox]

This PR makes the version of vLLM running accessible for operational maintainability via logfile and API.

• Logging - openai/api_server.py will log a header "vLLM API server version X.X.X" and llm_engine.py will log "Initializing an LLM engine (vX.X.X)" - This will help for forensics and startup validation.
• Version URL Route - This PR also adds /version as a route in the API server, with JSON output of vLLM version {"version": vllm.__version__}

So why add this? I have multiple openai API server instances running and need a hook to validate version consistency. I figure others would find this useful as well.

Example Logs

INFO 03-02 23:59:21 api_server.py:240] vLLM API server version 0.3.3
INFO 03-02 23:59:21 api_server.py:241] args: Namespace(host='0.0.0.0', port=8000, uvicorn_log_level='info', allow_credentials=False, allowed_origins=['*'], allowed_methods=['*'], allowed_headers=['*'], api_key=None, served_model_name='mistralai/Mistral-7B-Instruct-v0.1', lora_modules=None, chat_template=None, response_role='assistant', ssl_keyfile=None, ssl_certfile=None, root_path=None, middleware=[], model='mistralai/Mistral-7B-Instruct-v0.1', tokenizer=None, revision=None, code_revision=None, tokenizer_revision=None, tokenizer_mode='auto', trust_remote_code=False, download_dir=None, load_format='auto', dtype='auto', kv_cache_dtype='auto', max_model_len=20000, worker_use_ray=True, pipeline_parallel_size=1, tensor_parallel_size=1, max_parallel_loading_workers=None, block_size=16, enable_prefix_caching=False, seed=0, swap_space=4, gpu_memory_utilization=0.85, max_num_batched_tokens=None, max_num_seqs=256, max_paddings=256, disable_log_stats=False, quantization=None, enforce_eager=False, max_context_len_to_capture=8192, disable_custom_all_reduce=False, enable_lora=False, max_loras=1, max_lora_rank=16, lora_extra_vocab_size=256, lora_dtype='auto', max_cpu_loras=None, device='auto', engine_use_ray=False, disable_log_requests=False, max_log_len=None)
2024-03-02 23:59:23,441	INFO worker.py:1724 -- Started a local Ray instance.
INFO 03-02 23:59:24 llm_engine.py:88] Initializing an LLM engine (v0.3.3) with config: model='mistralai/Mistral-7B-Instruct-v0.1', tokenizer='mistralai/Mistral-7B-Instruct-v0.1', tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=20000, download_dir=None, load_format=auto, tensor_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto, device_config=cuda, seed=0)

Example APi - http://loalhost:8000/version

{
"version": "0.3.3"
}

[simon-mo]

Looks great!