WIP: Neaten/speed up examples

This is a first attempt to add an expected run time to the examples.

I reduced the number of live points in quite a few cases to make the run time faster. I also plan to update the docs to say that these won't produce really nice looking results and you need to bump up the settings to get that.

There are a couple of difficult cases.

• I couldn't get the custom likelihood pymc3 example to run, @matthew-pitkin would you be able to look at that? EDIT: I found a fix to this. The issue was that the custom prior couldn't be evaluated outside the sampler.
• I haven't installed multinest locally, I'll probably get around to that.
• I haven't run the open data examples yet, just because most of them are slow.

Oh, and I ran black on all the examples.