fix(data-warehouse): Use a seperate thread pool executor for warehouse pipeline

[Gilbert09]

Problem

• When we copy files from the local file system to S3 at the end of the DLT pipeline, the process blocks the main event loop and thus blocks the temporal heartbeater from beating - causing the pipeline to fail on large jobs
• Turns out using asyncio.to_thread uses the same default thread pool as the heartbeater

Changes

• Use a dedicated thread pool for the DLT pipeline so that the S3 file uploads don't block the heartbeater

[rossgray]

Looks good. To be honest I would have expected it to work fine before too especially since this is one of the suggested approaches in the Python docs. Though I have to admit a lot of asyncio stuff is a bit of a mystery and there's no better substitute to testing out in reality. So as long as you've tested it and it seems to fix the underlying issue then I'm happy with this change.