Function resolveChartText

The single definition of "the rendered text of this chart".

Priority order:

1. An emr_document with a type and a body renders through EMRDocumentFormatter. When its parsed content was rendered by an extraction, that body follows the formatter's header.
2. An Extend patient_chart upload renders through renderExtendChartFields — its structured extraction, NOT the OCR of the print, which is a single half-million-character blob we have no honest way to cut up. An upload that does not parse as an Extend payload still renders its notes/fields as JSON.
3. Anything left over falls back to raw_document.__circle_parsed_content.parsed_text.
4. Otherwise there is no text.

Parsed content is a fallback rather than the first choice on purpose. For an EMR document it is a snapshot of this formatter's output taken when the document was embedded, so preferring it would serve stale text whenever a formatter changes. It earns its place covering the cases the formatter cannot: uploads whose text was extracted from a PDF, and document types the formatter throws on.

Never throws — EMRDocumentFormatter.format throws on unsupported types and on malformed bodies, and one unrenderable document must not fail a whole batch.