Fixing batch sampling

[segasai]

This PR started as a fix to #415 , but because of the issue i thought we need to refactor that code, so I've now completely separated the creation of the batch_sampler in a distinct method that makes things much easier to understand. Also I've added more tests, some of which would have been tripped before the #415 fix

[coveralls]

Pull Request Test Coverage Report for Build 3596317313

• 122 of 127 (96.06%) changed or added relevant lines in 3 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.1%) to 91.729%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
py/dynesty/sampler.py	1	3	33.33%
py/dynesty/dynamicsampler.py	120	123	97.56%
<!--	Total:	122	127	96.06%	-->

Totals
Change from base Build 3559835188:	0.1%
Covered Lines:	4048
Relevant Lines:	4413

---

💛 - Coveralls

[segasai]

Any comments @joshspeagle, otherwise I'll merge and aim to push a quick release, as the performance issue is pretty substantial.

[joshspeagle]

Unfortunately, I don't have time for a detailed review on this over the next day or two, so I'd say just merge this in to go for the performance improvement (well, some return to some past performance + all the additional changes).

Thanks as always for continuing to clean up all the internals. The code is so much easier to work with now compared to when you first started taking a hammer to things.