Add `fail_open` usage metrics

Why are we doing this work

We want to better understand the usage of the recently introduced fail_open property for MR approval policies.

See: