Meld

This is a major version update of MELD, and it will break backwards compatibility with previous workflows due to a significant change in nomenclature accompanying our revised manuscript forthcoming in Nature Biotechnology.

You can find an up-to-date version of our article on BioRxiv with the revised language. The main difference is that we have dropped the "RES" and "EES" language for a more rigorous probabilistic interpretation inspired by an update to the algorithm made during the review process.

The output of `meld.MELD().fit_transform` is now referred to as the `sample_densities` and is the output of a kernel density estimation of each sample over that cell similarity graph. We then calculate the ratio of these densities using `meld.utils.normalize_densities` to calculate `sample_likelihoods`.

This framework reflects the interpretation of the MELD algorithm as a kernel density estimation over a graph. The tutorial and documentation have also been updated.

Fixes get_cmap import on init

This update implements the changes described in the revised version of the manuscript ([v3 on BioRxiv](https://www.biorxiv.org/content/10.1101/532846v3)).

* The meld.MELD() function now calculateds the EES using the new heat filter and provides a kernel density estimate of each sample over the data graph
* the meld.MELD() function can take in data and build a graph using the Graphtools package
* You only need to pass sample labels and the RES and cell number normalization happens automatically
* Updated parameters and comparison helper scripts

- Getting coverage to 100%
- Adding support for multidimensional RES
- Improving normalization in spectrogram clustering