Handle NA observations in the Illumina processor

### Context

Came up as a result of https://github.com/AlexsLemonade/refinebio/pull/2844

### Problem or idea

Our R code fails whenever an Illumina file contains an NA observation. More specifically, the failure that can't easily be fixed comes while we are doing background correction in `limma::nec`.

### Solution or next step

Figure out how common this problem is. I looked at ~60 experiments and only 1 or 2 of them had NA observations, so this problem might end up being relatively rare.

This problem could be alleviated by processing samples one-at-a-time instead of passing the whole dataframe to `limma::nec` at once. We need to investigate, however, if this has any adverse affects on our processing.

Start by locating the Illumina processor and its R background-correction path using limma::nec. Measure how often input files contain NA observations, then compare the current whole-dataframe processing with one-sample-at-a-time processing, including any adverse effects. Done means the prevalence and a safe handling approach are documented.

Written by the indexing model from the issue text.