Container supervisor for the per-scheduler nquic_server_sup partitions.

Holds N partition supervisors as children under one_for_one. Each child publishes its own pid to the listener's dispatch table on init, so a partition crash + restart is observed by nquic_dispatch:start_conn_child/3 on its next call without any republish step from this supervisor.

The partition count is published once into the dispatch table from this module's init/1 so the routing hash sees a consistent N.