update wp_scalogram demo work with matplotlib 2.0

[grlee77]

After upgrading to matplotlib 2.0, I get the following error in the wp_scalogram.py demo:

Traceback (most recent call last):
  File "wp_scalogram.py", line 43, in <module>
    ax2.specgram(data, NFFT=64, noverlap=32, cmap=cmap)
  File "/Users/lee8rx/anaconda/lib/python3.5/site-packages/matplotlib/__init__.py", line 1892, in inner
    return func(ax, *args, **kwargs)
  File "/Users/lee8rx/anaconda/lib/python3.5/site-packages/matplotlib/axes/_axes.py", line 7232, in specgram
    pad_xextent = (NFFT-noverlap) / Fs / 2
TypeError: unsupported operand type(s) for /: 'int' and 'NoneType'

If I specify Fs = 2, I get the same result as previously (e.g. y axis range from [0, 1.0]). I have no experience with the specgram function, so I am not certain if this is the best solution, but it seems to work.

here is the image from matplotlib 1.5.3:

and here is matplotlib 2.0:

note that the default interpolation changed to nearest and the new viridis default colormap is used

[grlee77]

looks like OS X on Travis CI took < 30 mins to start for this PR, so better than the 2 hours observed on the prior one (perhaps related to different load at differing times of day?)

[rgommers]

Wouldn't it be better to change the spectrogram plot back to linear interpolation. Don't have an installed version at hand to check, and reading the MPL source code is not too helpful with **kwargs everywhere, but does adding interpolation='linear' give back the more smooth plot?

[grlee77]

yeah, that works. I have changed it to interpolation='bilinear' to get a smooth plot again.

[rgommers]

LGTM now, merged. Thanks @grlee77