Add triu op (#5329)

* add triu

* add triu cuda kernel

* add triu gradient func

* add tensor test for triu

* add triu op

* add hard testcase for triu

* fix format

* add space

* add complicate test case

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>