Change AnalyzerChooser to not consider text

[ScottDugas]

This is a revert of #1613 which caused the index to use a different analyzer per record depending on the contents. This isn't really compatible with how IndexWriters are created, and hasn't been used.

I'm not entirely sure how it was intended to work, as you need some alignment with the query analyzer and the indexer analyzer. If you have two analyzers that behave differently for a given word, and then search for that word, how you analyze it, would depend on the surrounding contents of the field.

It also complicates allowing concurrent updates, since each record could in theory have different analyzers depending on the content, requiring the IndexWriter to be re-created.