Expose number of valid (non-nan) values

[nhey]

A bit cryptic, but this was already computed as Ns. It validates with no difference between python and opencl versions. Not tested against R, lmk if you want me to.

[mortvest]

Looks okay, but the opencl backend should return the number of valids even if the detailed_results is false.

[nhey]

Sure. Will fix.

[nhey]

Had some trouble recompiling futhark until I realised that the build was killed due to my system running OOM :P... I have regenerated bfastfinal.py using latest futhark master. Still validates.

Also I have included an unrelated commit (sorry), there was a check on self.detailed that caused an error when find_magnitudes was set to false. Corrected to self.detailed_results.

[mortvest]

Solid work, but you also need to add valids to the python-mp backend

[nhey]

I have amended the first and second commit to include the python-mp backend. It validates.

[mortvest]

Good work!