ensemble_weights.R review issues

[gaow]

The review is done by eyeballing so some comments may not be correct. Further AI assisted review might help:

1. We need to make sure Y aligns with each fold of data by sample name rather than assuming they follow default order --- at this point I am not sure how your Y connects to the cv_results output. If the connection is not water tight then perhaps we need to output some quantities from pecotmr's CV codes to ensure this match is fine?
2. I don't think it is such good idea to output some of the "performance metric" because these are in-sample metric and overfitted. in our CV results we output metric because they are out of sample and reliable. I would remove these metrics
3. The final weight is based only on the first data-set of input? That is unfair ... we should base it on the proper way in Dai et al considering all data-sets and all methods?
4. I think we should add a ensemble_twas_pipeline function to run the CV from pecotmr then combine with ensembe_twas_weights . There, we can provide default (cheap) ensemble method and compare that with the full set