using the compiled binary, add GPU compile and CI test, and fix the bug that "even if HTTP status code is not 200, it can pass the CI"

remove stdout print in endpoint.cpp

remove rocksdb compile, but source code remains temporarily

Revert "Accelerate rocksdb compile"

Accelerate rocksdb compile

Unify serve and web serve

update inference lib to 1.7.1 && support py3

update demo.gif