Collect HTML responses upstream of annotation

[josh-chamberlain]

Context

The tag collector compresses HTML responses into something we think will be useful. However, it's incomplete—we've decided to store the full HTML response and use that for training, perhaps still weighting based on header tags and whatnot. This gives us flexibility.

Requirements

• At the same time as we run the tag collector, upstream of annotation, store the full HTML response. This will stay with the data as it moves through the pipeline.