Add matrix multiplier methods

[mwtoews]

Add support for @ matrix multiplier methods. The Affine @ Affine operation (matrix-matrix) is the same as with *. Only the Affine @ tuple (matrix-vector) combination is slightly different, as the operator accepts an iterable of 2 or 3 values:

>>> from affine import Affine
>>> A = Affine(2, 0, 3, 0, 3, 2)
>>> print(A)
| 2.00, 0.00, 3.00|
| 0.00, 3.00, 2.00|
| 0.00, 0.00, 1.00|
>>> A @ (2, 3)
(7.0, 11.0)
>>> A @ (2, 3, 1)
(7.0, 11.0, 1.0)

and when it's 3 values, the third must value must be 1.0, because this is a 2D affine transformation module. Allowing 2 values enables a bit of backwards compatibility with the * operator, which only accepts 2 values.

I've added a note to enable PendingDeprecationWarning for a future release. I've held back in refactoring the examples in README.rst as often this is a user's reference for older releases.

Closes #107

[mwtoews]

I've just noticed (and corrected) this inconstancy with imul:

>>> from affine import Affine, identity
>>> t = Affine.translation(3, 5)
>>> t *= identity
>>> t
Affine(1.0, 0.0, 3.0,
       0.0, 1.0, 5.0)
>>> t *= (1, 2)
<stdin>:1: DeprecationWarning: in-place multiplication with tuple is deprecated
>>> t
(4.0, 7.0)

it's unexpected that an in-place operation would change the type to tuple, so I'm adding a warning.

The implementation of imatmul will allow another Affine, but raise TypeError with a tuple.

>>> t = Affine.translation(3, 5)
>>> t @= Affine.translation(2, 4)
>>> t
Affine(1.0, 0.0, 5.0,
       0.0, 1.0, 9.0)
>>> t @= (1, 2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/mtoews/src/affine/src/affine.py", line 604, in __imatmul__
    raise TypeError("Operation not supported")
TypeError: Operation not supported

[sgillies]

This is great @mwtoews !