process very large JSON files with jq using --stream/fromstream instead of loading the whole document into memory

Steps

1. Add --stream so jq parses the input incrementally and emits [path, leaf-value] / [path] event pairs as it goes, instead of waiting for the entire document to parse before producing output.
2. Reconstruct whole top-level values from the raw stream events using fromstream(...) combined with truncate_stream(inputs) at the desired path depth, e.g. `jq --stream 'fromstream(1|truncate_stream(inputs))'` to get one JSON value per top-level array element.
3. Use tostream (the inverse of fromstream) when converting an in-memory value back into streamed event form while composing custom stream-processing filters.
4. Filter events by path before reconstruction to avoid materializing parts of the document that aren't needed.
5. For newline-delimited JSON (one record per line), prefer plain `jq -c` processed line by line instead of --stream, since each line is already a small independent document.