fix bug in iswtn for data of arbitrary shape when using user-specified axes

[grlee77]

closes #460

The issue was a variable that should have only contained the shape of the transformed axes and not all axes. This was missed by the current tests because these used the same size on each axis of the transformed array so it didn't matter if the shape from the wrong axis was selected.

A new test case with various permutations of shapes and non-transformed axis was added. Specifically the test case tests the following shape, axes permutations:

shape = (1, 48, 32), axes = [1, 2]
shape = (1, 32, 48), axes = [1, 2]
shape = (48, 1, 32), axes = [0, 2]
shape = (48, 32, 1), axes = [0, 1]
shape = (32, 1, 48), axes = [0, 2]
shape = (32, 48, 1), axes = [0, 1]

[grlee77]

i would like to also include this one in the next bug fix release if anyone can do a quick review

[rgommers]

LGTM, merged. Thanks @grlee77

[rgommers]

that was a sloppy bug, makes me wonder if we should do odd shape testing in other functions

[grlee77]

makes me wonder if we should do odd shape testing in other functions

I think it is already reasonably covered for wavedecn/waverecn. See for example:

pywt/pywt/tests/test_multilevel.py

Lines 641 to 645 in 2e3a224

	def test_wavedecn_shapes_and_size():
	wav = pywt.Wavelet('db2')
	for data_shape in [(33, ), (64, 32), (1, 15, 30)]:
	for axes in [None, 0, -1]:
	for mode in pywt.Modes.modes:

and some other tests use odd size and/or a different size per axis.

[grlee77]

thanks for the quick review!