onnxruntime

Support int64 in ReduceMin cuda op for Opset 14 (#8307)

* reducemin int64_t support

* fix xxcuda.so load error

* testtest

* refactor

* update doc

* propagate types to opset14

* re-generate doc

* rename macro