Returns a Flow that implements a "brace counting" based framing stage for emitting valid JSON chunks. It scans the incoming data stream for valid JSON objects and returns chunks of ByteStrings containing only those valid chunks. Typical examples of data that one may want to frame using this stage include:

**Very large arrays**: {{{ [{"id": 1}, {"id": 2}, [...], {"id": 999}] }}}

**Multiple concatenated JSON objects** (with, or without commas between them): {{{ {"id": 1}, {"id": 2}, [...], {"id": 999} }}}

The framing works independently of formatting, i.e. it will still emit valid JSON elements even if two elements are separated by multiple newlines or other whitespace characters. And of course is insensitive (and does not impact the emitting frame) to the JSON object's internal formatting.