Refactor: Split 'do it' from reasoning/fallback

[alberth]

I randomly browsed some action server code, and eg 'find' and 'hand-over' does a lot more than just simply checking pre-conditions and (failing or) doing the requested thing.
It analyses what is missing, starts queries to get missing information, performs additional actions, etc.

I think it would be useful if a clean split is made between 'stupid' check & do-it code, and the intelligence/reasoning to make that elementary action smarter, so it can deal with more cases. Likely you'll end up with a tree of actions, where higher level actions use lower level actions to test and perform part of the action to perform.

@LoyVanBeek the reasoning code looks like it belongs in the conversation engine eventually.