Support state-dependent action space

Not sure if it is easy to support action spaces in this solver, but that would be a good modification for someone to do. It is not currently supported.

The issue does not name a file, test, or entry point. Start by locating where the solver assumes a fixed action space, then determine how actions should vary by state and what solver behavior would demonstrate support; the issue provides no concrete completion criteria.

Written by the indexing model from the issue text.