Non-contiguous matrices in triangular mul! and div!

[jishnub]

The LAPACK functions require contiguous matrices. Currently, the multiplication (BLAS) methods throw errors for non-contiguous arguments, while the LAPACK.trtrs! appears to silently accept a non-contiguous argument and returns an incorrect result (JuliaLang/julia#40613). This PR makes such cases fall back to the julia implementation, and only the stride-1 cases are forwarded to the corresponding LAPACK/BLAS functions.

Fixes JuliaLang/julia#40613
Fixes #878

After this, in-place multiplication by a triangular matrix works for non-contiguous matrices.

julia> U = UpperTriangular(float.(reshape(1:9,3,3)))
3×3 UpperTriangular{Float64, Matrix{Float64}}:
 1.0  4.0  7.0
  ⋅   5.0  8.0
  ⋅    ⋅   9.0

julia> B = float.(collect(reshape(1:25, 5, 5)));

julia> Bv_nc = view(B, 1:2:5, 1:2:5)
3×3 view(::Matrix{Float64}, 1:2:5, 1:2:5) with eltype Float64:
 1.0  11.0  21.0
 3.0  13.0  23.0
 5.0  15.0  25.0

julia> ldiv!(U, lmul!(U, Bv_nc))
3×3 view(::Matrix{Float64}, 1:2:5, 1:2:5) with eltype Float64:
 1.0  11.0  21.0
 3.0  13.0  23.0
 5.0  15.0  25.0

julia> rdiv!(rmul!(Bv_nc, U), U)
3×3 view(::Matrix{Float64}, 1:2:5, 1:2:5) with eltype Float64:
 1.0  11.0  21.0
 3.0  13.0  23.0
 5.0  15.0  25.0

[jishnub]

We may want to backport this change (at least the fix).