QR-Based Rank Determination and Pivoting

[dmbates]

• add statsrank to compute the rank and pivot vector for a model matrix using a pivoted QR decomposition
• modify tests, etc.
• remove statsqr and statscholesky functions

[palday]

@dmbates While I'm reviewing, can you go ahead and write a NEWS entry for this?

[codecov]

Codecov Report

Merging #479 (87f5554) into master (8d92489) will decrease coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #479      +/-   ##
==========================================
- Coverage   93.30%   93.28%   -0.03%     
==========================================
  Files          27       27              
  Lines        1912     1891      -21     
==========================================
- Hits         1784     1764      -20     
+ Misses        128      127       -1     

Impacted Files	Coverage Δ
src/Xymat.jl	85.29% <100.00%> (-1.55%)	⬇️
src/linalg/pivot.jl	100.00% <100.00%> (+2.94%)	⬆️

---

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 8d92489...cc3e996. Read the comment docs.

[palday]

I think this looks much cleaner. Should I try an MKL test run on the CIs?

[palday]

I like having this defense here (as a reminder to us as well), but note that @assert may disappear at some optimization levels.

[palday]

I've kicked off an MKL run

[dmbates]

I've kicked off an MKL run

I think it would be better to only do the MKL run on the Future platform. As you can see, it takes a long time to run on 1.5

[palday]

@dmbates The MKL run is now done on nightly.

When 1.7 is released, we can revisit making MKL part of the standard test suite instead of purely optional like it is now.

[dmbates]

@dmbates While I'm reviewing, can you go ahead and write a NEWS entry for this?

Done.