CI: Enrich for samples with TP53 and NF1 mutations that are in the stranded dataset

[jaclyn-taroni]

We need to account for the requirements to successfully run tp53_nf1_score when generating files in CI. Specifically - we can skip the poly-A and test only the stranded dataset (da8db99) but we have to make sure to do the following:

• Include positive examples in the SNV files (e.g., biospecimen IDs with TP53 and NF1 that are likely to be functional/pass the filtering here: https://github.com/AlexsLemonade/OpenPBTA-analysis/blob/a5dd6c122750fa145d8a60629f79ed856ee94c36/analyses/tp53_nf1_score/00-tp53-nf1-alterations.R) However - this is not enough (this didn't fix it: 18f335e).
• We also need select those biospecimens that have RNA-seq data in the stranded dataset - I might approach this by including a notebook in create-subset-files that looks into this and then hardcoding those values.

I'm going to comment that module out of CI until we make changes to create-subset-files so it doesn't block things. The alternative would be to include error handling that skips the ROC steps, but I think that puts us in a situation where we're not testing everything.