Linx.Reconcile.Source adapter for cgroup limits — lets the generic
Linx.Reconcile loop drive Linx.Cgroup.Reconcile.
The scope is the cgroup path (e.g. "/sys/fs/cgroup/myorg/web-42").
cgroupfs has no change-notification multicast, so subscribe/2 returns
:unsupported — the loop runs timer-only, which suits limit knobs that only
change when something writes them.
Reconciling the cgroup's limits is mechanism; its existence and membership are lifecycle the consumer's composite owns. A loop pointed at a not-yet-created cgroup simply reports the writes as failed until the composite creates it — it never tries to create it itself.
This is a thin delegation; for direct control use Linx.Cgroup.Reconcile.