Replace KnownIsNot* with *IsPlausible in pyhanabi.h.

[ghost]

This matches what pyhanabi.py was calling and what pyhanabi.cc was implementing.
It applies to both color and rank.

Also, change them (along with *WasHinted) to return bool instead of int since that's what the underlying functions return.

[NeilBurch]

Fix for ColorIsPlausible and RankIsPlausible already added separately: I looked at issues before pull requests. Thank you for also catching this.

Re. bool vs int, I would prefer to leave the use of int-as-bool in the CFFI interface: bool support in CFFI seemed both newer, and a bit rougher compared to support for int. For example, one supposedly up-to-date machine I tested on had cffi 1.9, which does not seem to support C++ bool.

Re. obs_dict, I would prefer to leave the RL interface untouched unless necessary, even if it might not be the most aesthetically pleasing choice. The extended hint information is exposed in both the vectorized observation and the pyhanabi observation. If I was to change the RL interface, my personal preference would actually be to remove almost all of the obs_dict entries as they're duplicates of the HanabiObservation object.