Add support for MLE and fractional n_components in PCA

Only explicit `n_components` parameters are currently supported but these extra behaviors could be useful to some, as in https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html.

I'm lifting this into an issue from an old comment at: https://github.com/dask/dask-ml/blob/b94c587abae3f5667eff131b0616ad8f91966e7f/dask_ml/decomposition/pca.py#L295-L302

Start with dask_ml/decomposition/pca.py at the referenced lines 295-302 and compare its current n_components handling with the scikit-learn PCA documentation. Determine how MLE and fractional values should behave in this implementation; done means both forms are supported consistently with the documented PCA behavior.

Written by the indexing model from the issue text.