improving the ack-filter

[dtaht]

@rmounce

I haven't got around to re-testing the final ack-filter code, but in my original tests ( http://blog.cerowrt.org/post/ack_filtering/ ) I'd felt that if all the ack drops counted towards the
DRR deficit, all 4 (instead of just one) flows would grow towards optimum faster as you'd
accumulate more acks to filter on each round.

I figure everyone's relaxed and enjoying summer, but I just wanted to write this idea down somewhere before I forgot it again.

[dtaht]

And indeed, I'd forgot this idea.