[DBA-137] Add the stop button

[23jura23]

Still a bit buggy, requires more testing

[gasparian]

I've added unittests for the query executor, feel free to revert back. It's just easier to understand what could go wrong here. Will try to generate e2e test, just cuious would it be possible to catch anything there)

[gasparian]

I've added unittests for the query executor, feel free to revert back. It's just easier to understand what could go wrong here. Will try to generate e2e test, just cuious would it be possible to catch anything there)

So I've decided to add a separate race condition tests as regular tests. Ptal if it makes sense. For now - everything pass.

[catstrike]

I don't think we should merge this in the current form.

The current approach terminates the thread directly, which is problematic because it can happen at an unexpected time.

Cancellation should be expressed at the API level instead. A better design would be for Thread.ask() to return a handle that exposes a cancel() method. The caller could then request cancellation without directly interfering with the thread itself.

The Executor already contains a loop that can handle cancellation gracefully:

for mode, chunk in compiled_graph.stream(
    start_state,
    stream_mode=["values", "messages"],
    config=config,
    **kwargs,
):

With a proper cancellation signal, this loop could exit cleanly without forcefully terminating the thread.

[SokolovYaroslav]

Lenar's concern is valid, however we'll probably move from streamlit, so it's not that critical