Is it possible to stream output?

Right now `execnb` is waiting for the final response from code execution. Is it possible to steam execution results without waiting for final response?

I'm using `execnb` for executing notebooks in the [Mercury](https://github.com/mljar/mercury) framework, and got such question from user https://github.com/mljar/mercury/issues/301

No file or test is named. Start by tracing execnb's notebook execution entry point and how it waits for the final response, then determine where incremental results could be exposed. Done means execution results are delivered before completion, with behavior demonstrated for the Mercury use case.

Written by the indexing model from the issue text.