mba-deploy-pod.sh EXIT trap cleanup silently fails on whey, 88 dirs/22G accumulated

nw-whey-cc found whey disk 95% CRIT: mba-deploy-pod.sh's TEMP_BASE clone (/tmp/mba-pod-<random>, ~260M each) accumulated 88 copies Sep30-Oct9 despite the EXIT trap at line 49 (rm -rf -- ${TEMP_BASE:?} 2>/dev/null || true). The || true swallows any rm failure silently - script exit code stays correct (per the 2026-09-30 fix note) but the dir is never confirmed gone. nw-whey-cc trashed all 88 manually to clear the CRIT; not their lane to fix the script. Fix: bypass any rm wrapper/alias on whey with 'command rm', and log a visible WARN on failure instead of swallowing it, without reintroducing the exit-code bug this trap was written to avoid.