Add timesteps to quadratic regularizers

[andgoldschmidt]

Modifies loss, gradient, and hessian of QuadraticRegularizer to use timesteps. That is, the loss is now 0.5 * (Δt .* vₜ)' * (R .* ( Δt .* vₜ)) instead of 0.5 * vₜ' * (R .* vₜ). This enforces an energy (integral) constraint, preventing the optimizer from stretching variable timesteps in order to minimize QuadraticRegularizer losses.

Closes #47

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (749a742) 49.67% compared to head (13b054e) 50.31%.
Report is 7 commits behind head on main.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #55      +/-   ##
==========================================
+ Coverage   49.67%   50.31%   +0.63%     
==========================================
  Files          17       17              
  Lines        2329     2359      +30     
==========================================
+ Hits         1157     1187      +30     
  Misses       1172     1172              

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