Add determinant(like scipy.linalg.det/slogdet) to dask.array.linalg

[lithomas1]

Just opening this issue to track missing bits from the array API missing in scikit-learn.

There's only one use of det/slogdet in scikit-learn, but I figured others might want to be able to do determinants in dask.

I think this can be implemented by doing a PLU decomp and then getting the determinant of the P matrix, and the product of diagonals of L and U.

Calculating the determinant of P (or more precisely its sign) doesn't seem to be easy, though.
(Maybe pivots, like scipy.linalg.lu_factor provides are needed?)