Code Review

This pull request refactors the communication between the xllm_service and backend xLLM instances from HTTP to RPC, a positive change intended to improve performance and simplify communication. However, a critical high-severity Denial of Service (DoS) vulnerability has been identified in the newly introduced HTTP header parsing logic. The GetJsonContentLength function in xllm_service/http_service/service.cpp uses LOG(FATAL) for missing headers and calls std::stoul on untrusted input without exception handling, which can lead to service crashes. Additionally, undefined behavior was found in the scheduler due to a missing return statement in a lambda function. Robust error handling is required to mitigate the DoS vulnerability, and the scheduler issue needs to be resolved.