[FEATURE] Update Corridor, ExploreMaze, BoxoHack to allow different actions

[ian-cannon]

🚀 Feature

Corridor, ExploreMaze, and BoxoHack each only allow for hard-coded actions. This change would allow users to define what actions they would allow in the environment.

Motivation

Transfer learning on these environments would be difficult if the action space is forced to change while it is allowed in other environments. This change would use the same method as other environments to allow new actions but still give defaults should the key not appear in kwargs

Pitch

Single line changes to each of the environments. I do not think these are large enough to warrant testing or API changes.