copytrito! for triangular matrices

[jishnub]

This does two things:

1. Forward copytrito! for triangular matrices to the parent in case the specified uplo corresponds to the stored part. This works because these matrices share their elements with the parents for the stored part.
2. Make copytrito! only copy the diagonal if the uplo corresponds to the non-stored part.

This makes copytrito! involving a triangular matrix equivalent to that involving its parent if the filled part is copied, and O(N) otherwise.

Examples of improvements in performance:

julia> using LinearAlgebra

julia> A1 = UpperTriangular(rand(400,400));

julia> A2 = similar(A1);

julia> @btime copytrito!($A2, $A1, 'U');
  70.753 μs (0 allocations: 0 bytes) # nightly v"1.12.0-DEV.1657"
  26.143 μs (0 allocations: 0 bytes) # this PR

julia> @btime copytrito!(parent($A2), $A1, 'U');
  56.025 μs (0 allocations: 0 bytes) # nightly
  26.633 μs (0 allocations: 0 bytes) # this PR