Broadcast operations for arbitrary styles

[jishnub]

In many broadcast operations that take an AbstractFill and produce another AbstractFill, the result should hold irrespective of the broadcast style involved. We may therefore specialize broadcast(op, F::AbstractFill) without caring about the BroadcastStyle of F.

E.g.:

julia> E = Eye(∞)
ℵ₀×ℵ₀ LinearAlgebra.Diagonal{Float64, Ones{Float64, 1, Tuple{InfiniteArrays.OneToInf{Int64}}}} with indices OneToInf()×OneToInf()

julia> E^2
ℵ₀×ℵ₀ LinearAlgebra.Diagonal{Float64, Ones{Float64, 1, Tuple{InfiniteArrays.OneToInf{Int64}}}} with indices OneToInf()×OneToInf()

This is a result that should hold irrespective of what the broadcast style of the diagonal is. Currently,

julia> Broadcast.BroadcastStyle(typeof(E))
LazyArrays.LazyArrayStyle{2}()

julia> Broadcast.BroadcastStyle(typeof(E.diag))
Base.Broadcast.DefaultArrayStyle{1}()

but if the style of E.diag is also changed to a LazyArrayStyle, the result of E^2 should remain unchanged.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.04%. Comparing base (877375f) to head (c2deb57).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #408      +/-   ##
==========================================
- Coverage   99.48%   99.04%   -0.44%     
==========================================
  Files           8        8              
  Lines        1157     1157              
==========================================
- Hits         1151     1146       -5     
- Misses          6       11       +5     

☔ 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.