fix(deploy_scale): dispatch to K8sDriver for a k8s-runtime fleet (#161)

deploy_scale/t_scale never got a k8s branch when studio#146 made the
read path (list/get/runtime_context) k8s-aware — start/stop on a
k8s-runtime fleet silently fell through to the ECS path with an empty
cluster string, which AWS's ECS API treats as the literal "default"
cluster, and with no fleet binding to match it, oab-mcp fell back to
the local machine's default AWS credential chain. Reproduced live:
stopping the "seaturtle" k8s test agent scaled a same-named ECS
service in the wrong AWS account instead.

t_scale now checks the named fleet's runtime first (mirroring
t_list/t_get/t_runtime_context) and calls the new
scp::scale_k8s_deployment (K8sDriver::scale) when it's k8s, before
ever touching the ECS cluster arg. The console's start/stop button and
the Tauri deploy_scale bridge now thread `fleet` through alongside
`cluster`, same as the existing list/runtime_context calls.