Run benchmarks nloop times

[RReverser]

Looks like at least some benches already used nloop variable in time calculations, but never looped. Each benchmark executing just once resulted in fairly unreliable numbers across runs.

This change makes each bench run 50 times, and averages the timing across them, making numbers more reproducible.

[kbarbary]

Is the runtime still pretty reasonable with nloop=50? I think maybe the reason I set nloop=1 was that, at least for the aperture functions, there's already an inner loop over the 1000 apertures.

[RReverser]

Is the runtime still pretty reasonable with nloop=50?

Yeah it's like 3-5 seconds on my laptop.

at least for the aperture functions, there's already an inner loop over the 1000 apertures

Ah okay, I was mostly looking at variations in sep.Background runs which wasn't covered by this. If you want, I can revert the changes to run the loop 50 times only for this function, but, as I said, it's still pretty fast anyway.