do_elastic_net: Adapted from...
In liangyy/mixqtl: Combine total and allele-specific counts for cis-QTL analysis

Fit y ~ x + 1 with elastic net using nested cross-validation. By "nested", we mean determine MSE using multiple partitions (K) instead of only one in vanilla cross-validation. The procedure is: 1. determine lambda sequence and fit y ~ x at different lambda values; 2. take in K copies of the partition of the input (one of the input, cv_fold_ids). For instance, 5-fold partition if you'd like 5-fold CV; 3. compute CV MSE at each lambda value for each fold in each partition; 4. compute CV MSE for each lambda as the mean of CV MSE of all partitions; 5. select the best model with the smallest CV MSE at that lambda.

1	do_elastic_net(cis_gt, expr_adj, n_folds, cv_fold_ids, alpha, ...)

`cis_gt`	x (genotype matrix, sample x variant)
`expr_adj`	y (expression level vector)
`n_folds`	cross-validation fold
`cv_fold_ids`	K partitions with N-fold (sample x K, where each column is partition of samples into N parts represented by the index of each part)
`alpha`	alpha parameter in glmnet
`...`	additional arguments passed to cv.glmnet