Reduce optimality gap for symmetric

[timholy]

It turns out that initializing with the solution to the unconstrained problem produces better results (in the sense of a closer match to quadratic optimality) than the previous diagonal initialization. Keep both as options, but default to the higher-quality one. The reason to keep the old strategy is that it is faster (about 2x on a 5x5 matrix), largely by avoiding the need to take logarithms.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (88c5b0a) to head (ff1e78b).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #10   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         3           
  Lines          226       258   +32     
=========================================
+ Hits           226       258   +32     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.