Remove Adjoint|Transpose(::UniformScaling) definition

[tkf]

Reading #25461, it looks like that the consensus was that Adjoint and Transpose have to add an extra layer of indirection by wrapping the given object. However, Adjoint(::UniformScaling) and Transpose(::UniformScaling) "eagerly execute" adjoint and transpose:

julia/stdlib/LinearAlgebra/src/uniformscaling.jl

Lines 66 to 69 in 91d2071

	transpose(J::UniformScaling) = J
	Transpose(S::UniformScaling) = transpose(S)
	adjoint(J::UniformScaling) = UniformScaling(conj(J.λ))
	Adjoint(S::UniformScaling) = adjoint(S)

(FYI: those are the added in commit e80697d#diff-2053e27c93f0362bfb456f36c77bb232 during #24969.)

If Adjoint(::UniformScaling) and Transpose(::UniformScaling) are just unnoticed during #25461, I suggest simply removing those lines (hence this PR). When I test stdlib/LinearAlgebra locally, those definitions were not necessary.

Note that, even though Adjoint{T, UniformScaling{T}} can be constructed with this PR, I'm not sure if it is necessary to add a special handling. Since Adjoint <: AbstractMatrix but not UniformScaling <: AbstractMatrix, an object of type Adjoint{T, UniformScaling{T}} is not well-defined; for example, after this PR, typing Adjoint(I) in REPL shows error from Base.show as would happen with other meaningless combinations like Adjoint("str").

[StefanKarpinski]

Makes sense to me. We've generally settled on Adjoint and Transpose always returning Adjoint and Transpose objects whereas adjoint and transpose can return something that is logically the adjoint or the transpose but not necessarily Adjoint or Transpose objects, e.g. by returning the argument in Hermitian or Symmetric cases or undoing a layer of Adjoint or Transpose wrapping.

[Sacha0]

Makes sense to me. We've generally settled on Adjoint and Transpose always returning Adjoint and Transpose objects whereas adjoint and transpose can return something that is logically the adjoint or the transpose but not necessarily Adjoint or Transpose objects, e.g. by returning the argument in Hermitian or Symmetric cases or undoing a layer of Adjoint or Transpose wrapping.

👍

Chances are these methods are vestigial. If not, I can't remember why :).

[Sacha0]

Thanks @tkf! :)

[tkf]

@Sacha0 Thanks for the review!