deferred drop, __parent workaround, reshape_fail, ACCT unbilling and some cosmetics

This set of patches addresses some issues _apparently_ not tightly bound
one to the other one. i tested the whole patchset in two configurations, one with cbq + sfq + ingress and the other with a packet scheduler that i'm actually working on. i'd be very glad if somebody volunteers for more testing, although i would be even more happy if somebody could read attentively and countercheck what i have written. i have checked and double checked slowly and carefully myself but i would really appreciate if somebody else does some further bug hunting, just in case i forgot something or i've made some wrong assumption.

Alessandro Salvatori
--
the _NOSPAM_ account is the one i am subscribed with, please remove _NOSPAM_ for personal replies