fix(models-http-api): use /completion and /embedding endpoint for llama.cpp

fix(models-http-api): use /completion and /embedding endpoint for llama.cpp #3787