tree.pyx: Fix and cleanup robinson_foulds() and related functions.

The main problem was that the properties used to compare were taken with
  getattr(node, attr)
but now they are normally accessed like
  node.props[prop]

And since you may need both (for example, in a PhyloTree, accessing
t.species can trigger a getter function, so it is not necessarily
the same as t.props["species"]. And thus all the fuzz with has_prop()
and get_prop().

Also, there is some cleanup and small simplifications, and a naming for
variables that better reflect what they mean (hopefully). Including
more changes attr -> prop ("attributes" being the old name, and
"properties" the new one).