How to handle sporadically failing downsampling tasks

Sometimes, downsampling tasks can fail. There are different ways to recover from this:

- use the CLI: `influx task retry-failed`
- query failed tasks in the system `_tasks` bucket and use the `/tasks/.../retry` endpoint to retry
- use a `from()` range 2-3 times as long as the `aggregateWindow()` so if a task run fails, it will recover the downsampled period the following task run