Allow non-float UpperHessenberg matrices

[dkarrasch]

Ideally, I would have liked to rewrite the 2-arg ldiv! and rdiv! into 3-arg versions to allow for unitful arguments (where source and destination have different units and where the former cannot be simply copied to the latter), but failed in a quick attempt. Anyway, this reduces the number of methods as follows:

EDIT: My original idea turned out to be nonsense. But I realized that the solve methods fail when the input matrix happens to be of non-float eltype. Probably rarely occurs, but also doesn't hurt to generalize.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.38%. Comparing base (f25276e) to head (631d46f).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1614      +/-   ##
==========================================
+ Coverage   94.37%   94.38%   +0.01%     
==========================================
  Files          35       35              
  Lines       16050    16050              
==========================================
+ Hits        15147    15149       +2     
+ Misses        903      901       -2     

☔ 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.