truncated_bytes() returns arbitrary values for empty streams #2887

Description

We are using simdjson::ondemand::document_stream in Apache Arrow's JSON parser.

When the input is empty and the stream contains no documents, truncated_bytes() can return a large arbitrary value such as 4294967295, causing empty input to be reported as truncated.

The behavior only occurs in the Arrow environment after repeated parser operations.

Question

Is truncated_bytes() supported when the stream contains no complete documents? If not, could this limitation be made more explicit in the documentation?

Context: apache/arrow#51038