Possible improvement to the way we are dealing with degenerate ellipsoids

[segasai]

I was thinking of changing the covariance matrix is 'improved' when it's invalid. As often the problem is that the ellipsoid is degenerate with eigen values being very small, a better fix should be keeping the eigen vectors and increasing the problematic eigen values to some level. This has the advantage of better preserving the shape of the ellipsoid in non-problematic dimensions. The code is less pretty though.

[coveralls]

Pull Request Test Coverage Report for Build 878850735

• 17 of 22 (77.27%) changed or added relevant lines in 1 file are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage decreased (-0.05%) to 71.86%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
py/dynesty/bounding.py	17	22	77.27%
<!--	Total:	17	22	77.27%	-->

Files with Coverage Reduction	New Missed Lines	%
py/dynesty/bounding.py	1	77.04%
<!--	Total:	1		-->

Totals
Change from base Build 873048010:	-0.05%
Covered Lines:	3381
Relevant Lines:	4705

---

💛 - Coveralls

[joshspeagle]

As often the problem is that the ellipsoid is degenerate with eigen values being very small, a better fix should be keeping the eigen vectors and increasing the problematic eigen values to some level. The code is less pretty though.

Agreed on both counts. The change also looks nice.