fix(gitlab-runner-gvisor): boot_verify pods die at their 1s deadline

boot_verify's synthetic job has no timeout, and FF_USE_POD_ACTIVE_DEADLINE_SECONDS (default on) turns the job timeout into the pod's activeDeadlineSeconds plus one: the canary pods got a one-second deadline and died two seconds after starting. Everything upstream of the deadline works now: auth, RBAC, scheduling onto the t2d fallback pool, image pull.

Disabling the flag for this service until the upstream boot_verify fix lands (runner MR follows). Job timeouts stay enforced by the runner, and the pod-cleanup TTL covers stragglers.

Tracking: gitlab-com/gl-infra/production-engineering#29646