Fix error with data ownership in cholesky

[craymichael]

Require that the fill-reducing permutation P of Cholesky factorizer owns its data before use. Fixes #271

This sets the OWNDATA flag of its data to True - in traceback it can be seen that col (line 102 of scipy/sparse/_index.py), which is derived from p, has this flag set to false. This creates a copy of the data if necessary (which apparently it wasn't in previous versions?)

Someone should give this patch a shot on their machine as well.

[codecov]

Codecov Report

Merging #288 (5e3f2e1) into master (b57b4cf) will increase coverage by 0.03%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #288      +/-   ##
==========================================
+ Coverage   95.21%   95.25%   +0.03%     
==========================================
  Files          22       22              
  Lines        3178     3223      +45     
==========================================
+ Hits         3026     3070      +44     
- Misses        152      153       +1     

Impacted Files	Coverage Δ
pygam/utils.py	87.46% <0.00%> (-0.27%)	⬇️
pygam/tests/test_terms.py	100.00% <0.00%> (ø)
pygam/pygam.py	94.80% <0.00%> (+<0.01%)	⬆️
pygam/tests/test_utils.py	96.55% <0.00%> (+0.04%)	⬆️
pygam/callbacks.py	95.23% <0.00%> (+0.07%)	⬆️
pygam/terms.py	94.60% <0.00%> (+0.19%)	⬆️
pygam/distributions.py	88.88% <0.00%> (+0.48%)	⬆️
pygam/tests/conftest.py	98.21% <0.00%> (+0.65%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b57b4cf...5e3f2e1. Read the comment docs.

[1tux]

this is amazing and saved my life!
I have to say that this Cholesky thing doesn't even improve running time.