Add missing matrix multiplication methods involving OneElement

[jishnub]

This adds several missing methods, so

julia> @btime OneElement(3, (3,3), (50,50)) * $(ones(50));
  8.797 μs (1 allocation: 496 bytes) # master
  70.977 ns (1 allocation: 496 bytes) # PR

julia> @btime OneElement(3, (3,3), (50,50)) * $(ones(50,50));
  138.330 μs (7 allocations: 50.42 KiB) # master
  1.254 μs (2 allocations: 19.61 KiB) # PR

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (9f5c6aa) 99.89% compared to head (77d751d) 99.90%.
Report is 5 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #347      +/-   ##
==========================================
+ Coverage   99.89%   99.90%   +0.01%     
==========================================
  Files           8        8              
  Lines         934     1037     +103     
==========================================
+ Hits          933     1036     +103     
  Misses          1        1              

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

[oxinabox]

This PR should hit 100% code coverage.
It wouldn't be too hard to add the 6 lines that are missed as identified by code-cov

[jishnub]

The codecov upload errored, e.g. in https://github.com/JuliaArrays/FillArrays.jl/actions/runs/7740101659/job/21104345601?pr=347. I think running the tests again might lead to more accurate results.

[oxinabox]

I am dubious as to if this check is worth the potential type instability.
I feel like such a OneElement where that element is also zero is almost never constructed.

[jishnub]

It's not type-unstable, though. It just calls a different method that returns the same vector, and should be type-stable as well. This is a guard against the current implementation allowing the indices to not lie within the axes (which probably should be disallowed).

E.g.:

julia> @report_opt mul!(zeros(2), ones(2,2), OneElement(1,2))
No errors detected

[oxinabox]

per above I am dubious

[oxinabox]

I am assume because the tests pass and this has 100% coverage that it is correct