[fix] revert reward update and evaluation (#6295)

* Revert "rewrite reward fn"

This reverts commit d06042b434396d5f2001e9540f308f932bd7477e.

* Revert "upgrade reward math verification"

This reverts commit a6085ff6765e7440bdcb94a823d8e8e88c1d2383.

* Revert "fix bug"

This reverts commit 01640ebd650baa173929743b1a609693b0380065.

* Revert "reuse comm-group"

This reverts commit bd61918dcfe7f4028f8ade4e8be351c4df991bf9.

* Revert "Support evaluation during training"

This reverts commit 57a88395feef9d3c4478fab126bdea8508308dbf.