local space (on python side)
We now export the mapper and methods to evaluate the basis function on the space. The approach is not very consistent. With the newest change, one gets a mapper object bound to an entity while 'evaluateBasis, javobisnBasis' are exported on the space taking an element.
A nicer option would be a 'localSpace' with a 'bind/unbind' method. That object could also be 'bound' to a quadrature and provide numpy storage to the basis set evaluation. We actually have the cache already on the python side which I believe is a raw piece of memory that we could perhaps map into a numpy array which could be reasonably efficiently used on the Python side.