Update to MOI v0.10

[ericphanson]

Closes #454
Closes #466

Seems like it's not too bad! Or I haven't gotten to the tests that will fail yet.

[codecov]

Codecov Report

Merging #467 (8648f21) into master (57e3f71) will decrease coverage by 0.54%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #467      +/-   ##
==========================================
- Coverage   92.36%   91.82%   -0.55%     
==========================================
  Files          83       83              
  Lines        5136     5137       +1     
==========================================
- Hits         4744     4717      -27     
- Misses        392      420      +28     

Impacted Files	Coverage Δ
src/solution.jl	84.18% <100.00%> (-8.44%)	⬇️
src/atoms/sdp_cone/lieb_ando.jl	88.57% <0.00%> (-11.43%)	⬇️
src/problem_depot/problems/sdp.jl	98.87% <0.00%> (-1.13%)	⬇️
src/variable.jl	95.37% <0.00%> (-0.93%)	⬇️
src/problem_depot/problem_depot.jl	68.75% <0.00%> (+6.25%)	⬆️

---

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 4fba10c...8648f21. Read the comment docs.

[ericphanson]

I think the GLPK failures (https://github.com/jump-dev/Convex.jl/runs/4200633319?check_suite_focus=true#step:6:1088) are GLPK issues, not Convex issues, since those tests pass with SCS; likewise, the nightly failure and SCS failure on the lieb ando problem isn't a Convex issue, since nothing there has changed. I'll update ConvexTests to try to surface these issues and file issues for the solvers.

I think this is good to go (assuming the docs build passes; at the time of writing it is still going, but the non-nightly tests have passed).

[odow]

Is that really it? I'm surprised. I'll make a new release of GLPK.

[ericphanson]

Yeah, the interface with MOI is fairly minimal, so I guess that has helped. That's since we do most things with our internal data structures and then emit a MBP-style conic problem, instead of using MOI bridges for more of our reformulations. Usually that's been a bad thing (since we can't support all the things that MOI could allow us to via bridges) but here I guess it's been convenient 😄.