world.plates->numpy, some optimizations

[tcld]

It seems I overlooked an array when making the other arrays numpy-arrays. I stumbled over this after noticing that center_world(), for some reason, returned different results for Python 2 and 3 when I generated the following test-world. That is fixed with this, too.

width=1024
height=768
plates=3
ocean=0.71
seed=13051

Short summary of the changes in here:

• Overlooked world.plates when switching to numpy.
• Added world.plates to the serialization-tests (It had been stored before, but never was tested, I think.)
• Reworked center_land() to make use of numpy.
• Removed some superfluous functions/tests.

PS: Sorry for still doing stuff that is not related to an immediate release!

[psi29a]

This actually looks like it should be included, because it falls under the mandate of 'use numpy, everywhere!'.

@ftomassetti wouldn't you agree?

[ftomassetti]

@psi29a I have nothing against it and the code seems ok to me, feel free to merge it!

[tcld]

Comments taken care of. I also removed another test + function since it wasn't used anywhere (and, like matrix_min_and_max(), had a numpy counterpart).

[ftomassetti]

Let's merge #198 and then this one before changing release number and tagging the new release