Add shape manipulation methods to ndarray

[sproshev]

All returned ndarrays should be like ndarray[T]

[shoyer]

It is valid to include only one dimension in reshape:

In [6]: np.zeros((2, 2)).reshape(-1)
Out[6]: array([0., 0., 0., 0.])

Zero entries seems to raise currently, but I suspect this is a bug -- it's probably safer just to make the overloaded signature *shape here

In [7]: np.zeros((1, 1)).reshape()
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-7-aca69783c5f4> in <module>()
----> 1 np.zeros((1, 1)).reshape()

(It's probably best to do the same for resize and transpose, too)

[sproshev]

Fixed!

[shoyer]

thanks!