Bot processes normally shouldn't take any significant CPU/disk and as little network as possible while the task is running.

That said, we have some workloads (ChromeOS build) that are so overwhelming that they can hang the bot processes (mainly task_runner) for several minutes, causing significant issues.

AI:
- Have the bot processes* run at a slightly elevated priority compared to the task priority, so  get a chance to send task updates.

In practice, only task_runner sends updates, so that's the primary point. Still, bot_main and run_isolated do cleanup in abnormal failure so they need to be able to get enough run time to be able to handle failure mode.