fix regularization in CRAIG

[amontoison]

I introduced a mistake in CRAIG with #104.

 [  γ    λ ] [ c₂   s₂ ] = [  δ    0 ]
 [  0    0 ] [ s₂  -c₂ ]   [  0    0 ]
 c₂, s₂, δ = sym_givens(γ, λ)

Or it's γ that we want to zero out and not λ.

 [  γ    λ ] [ -c₂   s₂ ] = [  0    δ ]
 [  0    0 ] [  s₂   c₂ ]   [  0    0 ]
 c₂, s₂, δ = sym_givens(λ, γ)

The commit fix this error, add a boolean parameter sqd + some tests for regularized problems and saddle-point / symmetric quasi-definite systems.

Docstrings have been updated too, it's Ax + λs = b that is solved when λ > 0 and not
Ax + √λs = b.

[codecov]

Codecov Report

Merging #171 into master will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #171   +/-   ##
=======================================
  Coverage   97.20%   97.20%           
=======================================
  Files          28       28           
  Lines        2503     2504    +1     
=======================================
+ Hits         2433     2434    +1     
  Misses         70       70           

Impacted Files	Coverage Δ
src/craig.jl	97.08% <100.00%> (+0.02%)	⬆️

---

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 d6c9bf4...c92b596. Read the comment docs.

[coveralls]

Coverage increased (+0.001%) to 97.204% when pulling c92b596 on amontoison:regularization-craig into d6c9bf4 on JuliaSmoothOptimizers:master.

[coveralls]

Coverage increased (+0.3%) to 97.199% when pulling 2072e5a on amontoison:regularization-craig into adc5bec on JuliaSmoothOptimizers:master.

[dpo]

Thank you. Here are a few simple comments. I'll have to check the other details carefully when I get back.

[amontoison]

I updated this pull request, it only concerns CRAIG now. It fixes the computation of a Givens reflection during the regularization.

[dpo]

Thank you.