High Variance Genes

v1.1 High Variance Genes (#53)

* Adds parameter n_top_var_genes
Allows specifying to only use genes with the most variance. The passed integer indicates how many to keep; all others discarded.
Default is 0, which keeps all genes.

* Various implementation improvements.
- Adds param `replace`: select sampling with or without replacement.
- Adds param `phenograph_parameters` as a single dict of keywords/values to be passed through to phenograph . If `k` is specified in `phenograph_parameters`, 'knn' is quietly ignored.
- Trims boost_rate to 0.5 when replace=True. Issues warning when necessary.
- Streamlined downsampling code thanks to `replace` parameter of np.choice.

* New library size method parameter updated.
- Adds parameter `new_lib_as`.
- Accepts method to be used directly. Method should accept a two element list of ints, and return a number coercible to int. Defaults to `np.mean`.

* Better pretty printing of results and other.