Drop pull-aware fatigue when puller has died mid-tick

[misterwise]

When a puller dies of TTL on the same tick a pull resolves, the move's body-weight fatigue could be stranded on the pulled creep instead of dying with the puller. Reproduction: a 2-MOVE puller pulls a 2-WORK harvester; on the puller's last tick, the puller is iterated first, movement.execute then the lifetime check calls _die and delete roomObjects[puller._id]. The pulled creep's later movement.execute enters _add-fatigue and the chain walk

while(!!object._pulled && !!roomObjects[object._pulled]) {
    object = roomObjects[object._pulled];
}

silently terminates because the puller is gone, then applies the move's fatigue to the pulled creep itself. With no MOVE parts to clear it the harvester is stuck forever (fatigue=4). Vanilla happens to produce the intended outcome when the pulled creep is iterated first, so the visible behavior was order-dependent.

Distinguish "no _pulled" (genuine chain head) from "_pulled but target missing" (dead puller). In the dead-link case return without landing fatigue anywhere — the move's pull-aware fatigue is buried with the puller it should have routed to. The "rest first" branch is preserved verbatim, so the negative-dFatigue / own-MOVE-clear path is unchanged.