Bounding update, getting rid of cholesky transform

[segasai]

To avoid forgetting this PR gets rid of cholesky transform when constructing ellipsoids. The reason is that since we are doing eigen transform anyway, we don't need to do cholesky on top. The transformation matrix from normal coordinates to ellipsoidal coordinates can be done with cholesky matrix L (where L L^T =C ) or through sqrt(L) V where (L is vector of eigen values and V are eigen vectors).

This speeds up the tests somewhat. On my machine it seems to be ~ 10%.

Also I relaxed one condition when checking whether the point belongs to multiple ellipsoids (in the hope of dealing with #396). Whether that is enough or not I don't know.

[coveralls]

Pull Request Test Coverage Report for Build 3170972253

• 13 of 19 (68.42%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.05%) to 90.467%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
py/dynesty/bounding.py	8	14	57.14%
<!--	Total:	13	19	68.42%	-->

Totals
Change from base Build 3088827015:	-0.05%
Covered Lines:	3891
Relevant Lines:	4301

---

💛 - Coveralls