cholesky(Positive, ...) does not play nicely with StaticArrays.jl

Hey, thanks for this great library!
We recently encountered that PositiveFactorizations.jl loses type information in case if input is a StaticArray.

julia> L = @SArray rand(3, 3)
3×3 SMatrix{3, 3, Float64, 9} with indices SOneTo(3)×SOneTo(3):
 0.0501647  0.544364  0.896714
 0.243666   0.328148  0.929278
 0.755197   0.661655  0.979631

julia> A = L'L
3×3 SMatrix{3, 3, Float64, 9} with indices SOneTo(3)×SOneTo(3):
 0.632212  0.606946  1.01123
 0.606946  0.841801  1.44126
 1.01123   1.44126   2.62733

julia> inv(cholesky(A)) # Original Base function returns StaticArray
3×3 SMatrix{3, 3, Float64, 9} with indices SOneTo(3)×SOneTo(3):
  5.27106    -5.37836    0.921597
 -5.37836    25.0268   -11.6587
  0.921597  -11.6587     6.42145

julia> inv(cholesky(PositiveFactorizations.Positive, A))
3×3 Matrix{Float64}:
  5.27106    -5.37836    0.921597
 -5.37836    25.0268   -11.6587
  0.921597  -11.6587     6.42145

Not sure if it is easy to fix, but would be great to make it work with StaticArrays as well.

Start by reproducing the reported cholesky(Positive, A) behavior with the StaticArrays example, then inspect the PositiveFactorizations cholesky and inv entry points. Done means the positive-factorization path preserves the StaticArray result type like Base cholesky does, with a regression check for the shown case.