Have `paduapts` return the data as the transpose of the current format?

Right now paduapts(n) returns an N x 2 matrix. I'm proposing that it instead return a 2 x N matrix. The reason is that this allows for a no-op conversion to a Vector{Vec{2,Float64}}: the current code works

pts=paduapts(10)
ptr=reinterpret(Ptr{Vec{2,Float64}},pointer(pts'))  # reinterpret the memory of pts'
unsafe_wrap(Vector{Vec{2,Float64}},
            ptr,
            size(pts,1),true)  

With the proposed change, the pts' would become pts and no new memory would need to be allocated.

Start at the paduapts(n) implementation and inspect callers that rely on its current N×2 shape. Confirm that the proposed 2×N return shape preserves the point data and update any affected expectations or tests; done means paduapts returns the new format without unnecessary allocation.