Add Replay Buffer for MAGRPO, MAAC, IAC, and use env_step to record (#50)

* add rolloutbuffer for magrpo to align with maac

* Update requirements.txt

* Use env-step logging for MAAC/MAGRPO

* allow verbose output

* update default

* Update magrpo.py

* Update magrpo.py

* Revert "Update magrpo.py"

This reverts commit e0ce522.

* Revert "Update magrpo.py"

This reverts commit c665345.

* Reapply "Update magrpo.py"

This reverts commit bdb822b.

* fix magrpo

* remove hard constraint of buffer size of maac

* try to align magrpo with maac by having top-k null and same prompt

* update magrpo log

* ud wandb's problem

* fix logging

* Revert "fix logging"

This reverts commit 2a1653b.

* Revert "ud wandb's problem"

This reverts commit 1a27c31.

* Revert "update magrpo log"

This reverts commit e99796b.

* get Iac setup with buffer and remove unused return norm in maac

* add eval to iac

* allow multi-turn iac

* update log of iac

* better align

* Update iac.py

* Update changelog.md

Author: LovelyBuggies