Support graceful and quick farmer shutdown in the middle of plotting process

[nazar-pc]

There were a few reports about this and it annoyed me as well.

Essentially what was happening here is a blocking operation of plotting that outside select couldn't interrupt until sector plotting finishes, which in case of network issues can take a very long time.

Here we do two things:

• since plot_sector is async by itself, we race it with shutdown notification
• inside of plot_sector we additionally yield to the async executor after each piece processed such that plotting can be interrupted in the middle of that compute-intensive process as well, not just during async piece downloading

Node that there are no yields when sector is ready and is being written to disk because this is both wasteful and shouldn't take a lot of time anyway.

Review with whitespace changes ignored, the diff is not that large.

Code contributor checklist:

• I have read, understood and followed contributing guide