check_and_purge_safe_mem livelocks forever when the head allocation fails _free_safe_mem's integrity checks — unbounded kernel loop + tbridge_printf spam turns any detected corruption into a hard hang

Summary

The purge loop's only exit is the head becoming NULL, but the free's three error paths return without unlinking - a corrupt head makes every iteration re-read the same node and spin forever printing SAFEMEM BUG lines (the corruption detector becomes the hang). Dead code in base (zero in-tree alloc/free/purge callers; root-gated module; trigger requires a prior DF-2156/2157/2158 manifestation). Return-int + skip-bad-node fix.