Avoid potential ambiguities for [h/v/hv]cat in packages

[dkarrasch]

#41394 has changed the method signature in concatenation functions involving UniformScalings. This triggers method ambiguities for matrix-related packages, including LinearMaps.jl (see JuliaLinearAlgebra/LinearMaps.jl#158). This PR resolves the ambiguity at the root.

Closes JuliaLinearAlgebra/LinearMaps.jl#158.

[pablosanjose]

Friendly bump [I need this! :-)]

[chriscoey]

I do too 🙂

[andreasvarga]

This would be very helpful for me too. See my last succesfull tests with DescriptorSystems with Julia 1.6 and the 174 failed tests with Julia 1.7 due to the [h/v/hv]cat issue.

[KristofferC]

Seems like this would be useful to people and is unlikely to have any real issues so I would say to just merge this.

[dkarrasch]

The issue is that the PR causing the ambiguity (#41394) has introduced heavy latency problems, so I think triage needs to discuss/decide whether we want to go forward merging this or go backward and revert #41394.

[DilumAluthge]

I'm removing the merge me label so that this can be discussed at the next triage call.

On a separate note, I restarted the two failing CI jobs, just so we can see if CI passes on this PR.

[DilumAluthge]

Actually, I'll add the merge me label back.

If triage decides to revert #41394, but this PR has already been merged, presumably we can just revert this PR first and then revert #41394 (or revert them both in a single PR).

[JeffBezanson]

If this PR resolves the problem, I'm happy, but if anybody knows a remaining reason to revert #41394 speak up.