Nightly backup of the self-hosted Git service was failing

The nightly backup had been failing every night since the service moved to replacement hardware.

The backup jobs themselves had been updated to point at the new machine, but the schedule that runs them still named the old one, and that outer setting overrides the job. Every run therefore resolved to a machine that no longer exists and stopped before doing any work.

The schedule now follows the jobs, so there is a single place to change if the service moves again. Verified by a successful run against the current host.