Matrix conversion for structured matrices without a zero

[jishnub]

This adds a fallback branch in case the eltype doesn't have a unique zero, so that the following works after this PR:

julia> D = Diagonal(fill(ones(2,2), 2, 2))
2×2 Diagonal{Matrix{Float64}, Vector{Matrix{Float64}}}:
 [1.0 1.0; 1.0 1.0]          ⋅         
         ⋅           [1.0 1.0; 1.0 1.0]

julia> Matrix(D)
2×2 Matrix{Matrix{Float64}}:
 [1.0 1.0; 1.0 1.0]  [0.0 0.0; 0.0 0.0]
 [0.0 0.0; 0.0 0.0]  [1.0 1.0; 1.0 1.0]

Similarly, for Bidiagonal and Tridiagonal/SymTridiagonal.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.96%. Comparing base (62c8100) to head (4bc7eda).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1159   +/-   ##
=======================================
  Coverage   91.96%   91.96%           
=======================================
  Files          34       34           
  Lines       15473    15485   +12     
=======================================
+ Hits        14229    14241   +12     
  Misses       1244     1244           

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

[dkarrasch]

LGTM. My only question is (as in the other PRs) about implicit eltype promotion. Would that work as well?

[jishnub]

The problem with this approach is #1221