Report Oban job failures to Sentry only after all retries are exhausted

Hive now waits until an Oban job exhausts all of its retry attempts before reporting the failure to Sentry. Previously, a first-attempt timeout (for example in the GitHub issue classification worker) generated a Sentry alert even though Oban would retry the job and only discard it after the final attempt, creating alert noise on transient failures. The new default keeps the worker behavior unchanged and moves the noise control into Sentry’s Oban integration, which now reports only when the attempt number reaches max_attempts. Operators who want to be alerted on every failed attempt can opt back in by setting SENTRY_OBAN_REPORT_RETRIES=true. The same quieter default is also applied through the Helm chart values, and no migration is required for existing deployments.