pycbc_inspiral backend test in CI

[GarethCabournDavies]

(partly just thinking out loud)

I don't think this exists, but I think that we should have a CI test for which pycbc_inspiral (and possibly others) run on the same data using different FFT backends, maybe on the CPU and GPU artchitecture as well, and check that output is comparable. My thinking is:

1. Check that all FFT backends work
2. Check that we don't introduce bugs(features) which don't(only) affect the path but not(only) on the one which is being tested by the CI

The outputs won't be the same due to the way the FFT libraries work, so:

• tolerances will need to be defined with careful consideration
• Is there a way that these could be forced to be identical?