More reasoning and planning algorithms beyond the defaults

[farice]

As we move towards a general tool for prototyping LLM reasoning and planning algorithms, we should decouple the notions of:

• algorithm being prototyped
• benchmark / problem set being visualized