[Feature] Fleet update runs continue on failure based on setting of acceptable #/% cluster failures

As a fleet administrator I want define the number of cluster failures an update run can encounter before it stops so that my update run will still continue even when I have some problematic clusters.

Criteria:

- The default behavior is to accept no failures (current behavior).
- I can set a threshold for acceptable number of cluster upgrade failures as either an absolute number or percentage.
- I can set a threshold for the entire run, or for more granular scopes (stages, groups)
- On reaching threshold update run should halt with explicit error message.