Use more efficient matrix vector multiplication op in proposals

[matt-graham]

#47 switched to always using dense matrix constructed with diag for diagonal shapes, which incurs a quadratic cost matrix vector multiplication operation, even though it should be possible in linear, but this was found to much quicker in profiling than using Matrix::diagonal. This PR instead defines a custom matrix-vector multiplication operator %@% to use in proposals that treats vector shapes as diagonal matrices and scalar shapes as scaled identities, and performs the matrix vector multiply in linear cost.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (71f0af1) to head (5c4195b).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #59   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           11        11           
  Lines          518       520    +2     
=========================================
+ Hits           518       520    +2     

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