feat: add inference backends

[ChenZiHong-Gavin]

This PR introduces several LLM api client and inference backends, including:

1. http_client
2. ollama_client
3. openai_client
4. hf
5. sglang
6. tgi(wip)
7. trt(wip)
8. vllm(wip)

.env changes can be found in .env.example

# Tokenizer
TOKENIZER_MODEL=

# LLM
# Support different backends: http_api, openai_api, ollama_api, ollama, huggingface, tgi, sglang, tensorrt

# http_api / openai_api
SYNTHESIZER_BACKEND=openai_api
SYNTHESIZER_MODEL=gpt-4o-mini
SYNTHESIZER_BASE_URL=
SYNTHESIZER_API_KEY=
TRAINEE_BACKEND=openai_api
TRAINEE_MODEL=gpt-4o-mini
TRAINEE_BASE_URL=
TRAINEE_API_KEY=

# # ollama_api
# SYNTHESIZER_BACKEND=ollama_api
# SYNTHESIZER_MODEL=gemma3
# SYNTHESIZER_BASE_URL=http://localhost:11434
#
# Note: TRAINEE with ollama_api backend is not supported yet as ollama_api does not support logprobs.

# # huggingface
# SYNTHESIZER_BACKEND=huggingface
# SYNTHESIZER_MODEL=Qwen/Qwen2.5-0.5B-Instruct
#
# TRAINEE_BACKEND=huggingface
# TRAINEE_MODEL=Qwen/Qwen2.5-0.5B-Instruct

[tpoisonooo]

1. 对应的 README 要更新，支持哪些 feature 得能看到
2. 还有个有效的是自动扩展负载， 例如调一下 ray[serve] api

https://github.com/SeedLLM/DataPolisher/pull/1/files#diff-5648623a11374bdc84a573cac0a89d4e93d162c80c8938c82780f76c96c4373c

[tpoisonooo]

这种：

或者这种：

都可以参考。