Detect Affymetrix HGU133A-HGU133B "pairs" in an expression data-driven manner (automatically)

### New Issue Checklist
- [x] The title is short and descriptive
- [x] The issue contains an:
- [x] Idea (new feature, user story, etc)
- [ ] Problem
- [x] You have explained the:
- [x] Context
- [x] Problem or idea
- [x] Solution or next step

### Context

_Explain the conditions that led you to write this issue. If you are proposing a new feature, the context should be your user story. If this relates to a certain page or API endpoint, provide a link._

Older Affymetrix chips -- `hgu133a` and `hgu133b` -- sampled a smaller number of probes/genes. Sometimes samples would be run on both `hgu133a` and `hgu133b` to get better coverage of the transcriptome (sort of equivalent to running a sample using the later `hgu133plus2` platform) resulting in paired `a` and `b` samples.

### Problem or idea

_The context should lead to something, an idea or a problem that you’re facing._

It would be great to automatically detected these paired samples and combine the information from both chips. Prior to taking this approach, there will be a considerable number of missing values in these samples as compared to their genome-wide successors. Some kind of imputation approach may be the path forward. If we use a neural network-based approach to reconstruct the other "half" of the sample, we may find that the true paired sample is most highly correlated with our reconstructed values.

### Solution or next step

_You can tag others or simply leave it for further investigation, but you must propose a next step towards solving the issue._

It's fine to run SCAN on these samples as planned. We should investigate this further after we have processed a sufficient number of samples to design experiments for investigating this issue.

Here's an example experiment from ArrayExpress: https://www.ebi.ac.uk/arrayexpress/experiments/E-GEOD-11908/

No files, tests, or code entry points are identified in the issue. Start by reviewing the existing SCAN processing and the ArrayExpress E-GEOD-11908 example, then define experiments for detecting paired hgu133a and hgu133b samples and evaluating imputation; the work is done when an agreed automatic approach is validated on processed data.

Written by the indexing model from the issue text.