swap_pager_putpages stripe-boundary trim is off-by-one, invoking freeswapspace with npages=0

Summary

swap_pager_putpages trim check at :1713 (blk ^ (blk+n)) & ~SWB_DMMASK fires when blk+n lands exactly ON a stripe boundary even though the I/O itself does not cross. Computed trim sets j=n -> swp_pager_freeswapspace(object, blk+j, n-j=0) -> blist_free(swapblist, X, 0). blist code not designed for count=0; updates bm_bighint spuriously on fully-allocated meta nodes -> suboptimal allocation hints + minor fragmentation. No panic in practice (leaf-level shift UB unreachable). Should use blk+n-1 (last page in I/O). Fires ~1/64 per single-page putpages cluster.