Change from resampling to use pyFFTW

[calum-chamberlain]

What does this PR do?

Changes the "resampling" process in utils.pre_processing to use .interpolate rather than .resample for speed and stability.

Borrows code from obspy for resampling, using the pyfftw interfaces in place of scipy fftpack.

Why was it initiated? Any relevant Issues?

311

PR Checklist

• [x] develop base branch selected?
• [x] This PR is not directly related to an existing issue (which has no PR yet).
• [x] All tests still pass.
• [x] Any new features or fixed regressions are be covered via new tests.
• [x] Any new or changed features have are fully documented.
• [x] Significant changes have been added to CHANGES.md. - [ ] First time contributors have added your name to CONTRIBUTORS.md.

[calum-chamberlain]

This is going to be a bigger job than anticipated. Closing this but might come back to it at some point.

[calum-chamberlain]

I decided to try just using pyFFTW plugged in to resampling - this ends up being a lot faster than using scipy fft routines for large primes, and faster than interpolation. It is also a simple plugin for this code, and we already rely on FFTW for correlations, so installation doesn't change much (with FFTW installed pyFFTW can be installed via pip, or conda - conda will install FFTW itself if needed).