feat: use isoform lengths median to determine gene feature_length + write feature_length for 'spike-in' features in var

[nayib-jose-gloria]

Reason for Change

• cellxgene-schema CLI must update annotation of var['feature_length'] #990

Changes

• changed the calculation of gene length from "merged" to "median" using GTFtools implementation.
• we previously set feature_length to 0 for rows in var with feature_type "spike-in", despite having the feature length pre-calculated. We will now write the feature_length in write-labels for all feature types; this command is run after an H5AD is uploaded + validated in the CxG data portal
• ran gene_processing.py and generated new gene metadata files (genes_*.csv.gz) using new feature-length calculation. these files are the source of truth for the feature-length annotated onto H5AD var df after upload.

Notes for Reviewer

• no changes to genes_ercc.csv.gz due to always having a different method for calculating feature length, unrelated to above
• no changes to genes_sars_cov_2.csv.gz due to each gene in that file only having 1 isoform in gtf, so no changes between merged vs. median calculation

[sidneymbell]

This implementation LGTM! Very tidy. Thank you!

Note that I haven't had time to review the output files in detail, but the code looks good.