Fix color masking in Pixie

[camisowers]

If you haven't already, please read through our contributing guidelines before opening your PR

What is the purpose of this PR?

Closes #1117. Fixes colored masking for both pixel and cell clustering.

How did you implement your changes

The problem seemed to arise from non-sequential meta_cluster ids. If you change meta cluster 2 -> 21, then the IDs would be [1,3,4, ..., 21], which causes an issue when creating a colormap that expects only 20 clusters (and therefore values up to 20). However, upon further investigation, most of the cell cluster colors were being visualized incorrectly.

To ensure we are always matching the clusters correctly no matter the meta_cluster ID assignments, we

1. Reassign regular (non-colored) mask values to be sequential ints and save it to the *_meta_cluster_mapping.csv for reference. Use sort_values(by=['cell_meta_cluster']) to ensure the cluster_id are generated by ascending meta cluster number every time for consistency.
   • The cell masks already did this, but it was not using the saved cluster_id values in *_meta_cluster_mapping.csv and was instead doing it's own (second) reassignment, which was why the mapping was wrong.
2. When generating colored masks, simply go into the saved file so we can grab the cluster_id that corresponds to the meta_cluster number, and combine that with the raw_cmap (e.g. {meta_cluster: color}), so we can know for sure which mask integer ID connects to each color.

Old (no good) colored mask output:

New (correct) colored mask output:

Verified by me hand checking the label and the provided cell_meta_cluster_rename in the cell table for a bunch of random cells in the image. 😅

Remaining issues

• Check pixel clustering mantis mapping file.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[cliu72]

A few suggestions!

[cliu72]

Looks good to me!

[alex-l-kong]

Looks good

[srivarra]

Wild stuff!