Support array eltypes in zeroeltype for Mul getindex

[dlfivefifty]

Indexing a Mul (_getindex) starts from zeroeltype(M) = zero(eltype(M)). That fails when the entries are arrays without a type-level zero, e.g. Vector{Float64} or Matrix{Float64}:

julia> X = reshape([[1 2; 3 4], [5 6; 7 8]], 1, 2);

julia> ArrayLayouts.Mul(X, [1, 2])[1]
ERROR: MethodError: no method matching zero(::Type{Matrix{Int64}})

For Mul with an AbstractArray eltype, this PR takes the zero from the entries instead: zero(first(M.A) * first(M.B)). Other eltypes keep zero(eltype(M)), and existing specialisations of zeroeltype are unaffected.

This came up with sums of matrix-valued expansions in ClassicalOrthogonalPolynomials.jl, e.g. sum([exp(x) cos(x); sin(x) 1] for x in 0..1).

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.23%. Comparing base (1e4fcc1) to head (39f4e97).
⚠️ Report is 11 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #280       +/-   ##
===========================================
+ Coverage    0.00%   90.23%   +90.23%     
===========================================
  Files          11       12        +1     
  Lines        1858     1966      +108     
===========================================
+ Hits            0     1774     +1774     
+ Misses       1858      192     -1666     

☔ View full report in Codecov by Harness.
📢 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.