Do not densify the sparse arguments of a broadcast + or - whose result is empty

[ViralBShah]

Fixes a regression from #1004, found in review: a broadcast + or - of a sparse array with a dense array converted the sparse arguments to Arrays before looking at the shape of the result, so an empty result cost memory proportional to the size of the sparse argument, and could run out of it.

Before

julia> x, z = spzeros(1_000_000), zeros(1, 0);

julia> size(x .+ z), @allocated(x .+ z)
((1000000, 0), 8011936)

After

julia> size(x .+ z), @allocated(x .+ z)
((1000000, 0), 48)

_densebroadcast now hands the broadcast to the dense kernel as it is when one of its axes is empty: that kernel reads no entry, so the reason to densify (generic broadcast indexing the sparse arrays entry by entry) does not apply. The result type is the same on both paths and stays inferable.

Two assertions that #1004 and #1005 added are removed as duplicates of ones in the older loops of test/higherorderfns.jl: broadcast(+, S, d) (the sums of a sparse matrix with a dense vector in the mixed-argument loop) and cos.(view(S, :, [2, 1])) (the loop over views of sparse arrays, which has a view of columns picked by a vector).

Not a backport candidate: #1004 is on no release branch.

Written by Claude Code.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.59%. Comparing base (26dce7e) to head (a12b474).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1007   +/-   ##
=======================================
  Coverage   95.59%   95.59%           
=======================================
  Files          16       16           
  Lines        9304     9304           
=======================================
  Hits         8894     8894           
  Misses        410      410           

☔ 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.