add specific Adjoint{StridedMatrix} * SparseVector method

[dkarrasch]

Closes #32402.

A quick adoption of the existing Transpose method, plus some minor code simplification/consistency.

[dkarrasch]

Test failure on windows is in download.

[ViralBShah]

Do we have adequate test coverage of these cases?

[dkarrasch]

Do we have adequate test coverage of these cases?

What do you mean? We had tests for denseMat*sparseVec and transposeDenseMat * sparseVec. I added tests to cover Adjoints and added Complex eltypes where Transpose and Adjoint make a difference.

[ViralBShah]

Looks good to me.

[dkarrasch]

If anything, then I see the "issue" of code duplication. I see that we usually don't make much effort in code reduction in the area of matrix multiplication, because of the risk of method ambiguity. But I guess if code reduction is desired this could be another PR.

Just to give an example, the 3-arg mul! simply forwards its arguments to the 5-arg mul! by appending one and zero. Currently, we have 3 individual methods for that, for StridedMatrix, Adjoint and Transpose. Other reductions include the trick we sometimes use in tests, for-loop over transform in (adjoint, transpose) and then @eval and the function definition etc.

[ViralBShah]

I would be happy to do code reduction separately and get performance improvements in for now.