ENH: add num_interp to AlignedSpin prior

[mj-will]

#849 added a new test to the prior_test.py but this test is currently taking ~1 hour total in the CI. To address this, I've added num_interp to the AlignedSpin prior.

This allows us to speed up the test but allows the user to configure the interpolation. The defaults are the same (different for each case).

On my laptop, this reduced the time from ~20 minutes to ~ 2 minutes, but we'll see how the CI behaves.

[adivijaykumar]

I haven't thought too much about this, but will the default of 10_000 and 100_000 also make sampling slower? We want to avoid that too, right?

[mj-will]

I haven't thought too much about this, but will the default of 10_000 and 100_000 also make sampling slower? We want to avoid that too, right?

So neither should affect sampling, it should be a one-time cost. Even with 100,000 the simple method seems to be quick enough that it doesn't slow down the tests.

Though perhaps it would be good to document these choices, @ColmTalbot was there a particular motivation behind these choices?

[ColmTalbot]

I'm fairly sure the original reason for bumping up the number is accuracy close to the peak. The density diverges at zero and so to resolve this well we need a lot of points. I think this is a good compromise.

[mj-will]

Thanks both, I'll probably add this in to 2.5.0 just to make preparing the release slightly quicker.