New registry for envs reward functions

### Description

Introduced in #137 .

Create an empty dict in the BaseEnv to collect all rewards functions. Introduce somekind of StrEnum per env implementation to dynamically select functions to call.

### References

_No response_

Start by reviewing issue #137 and the BaseEnv, then inspect the environment implementations and how their reward functions are currently selected. The work is done when BaseEnv has an empty reward-function registry and each environment implementation has a StrEnum-based way to select the functions to call.

Written by the indexing model from the issue text.