Gracefully handle situations when there are subsets of "active samples" that do not have data from the given assay_name under test

[lianos]

For instance, you may have defined a linear model via flm_def() over a set of samples, and then wanted to fit/test that model using one of the assays in the data set, like assay_name = "rnaseq" for instance.

There is no guarantee that all of the samples used to create the design matrix will have been measured with the "rnaseq" assay, and the biocbox returned here won't be concordant with the samples you sent in to retrieve the assay matrix, and therefore will fire an error here.