tor-congestion: WIP implementation of RTT estimation

This tries and fails to implement the round-trip-time estimation algorithm in Prop#324 ("fails" meaning "doesn't pass the test vectors right now"). There's a clock estimate test which seems flawed (since the structure of my code and the C code is different), and there's an RTT test which is broken for reasons I don't understand.

cc @dgoulet: if you could take a quick look through the algorithm and see if there's anything obviously wrong here (or cc someone else who can), that'd be very appreciated! (Otherwise I'll probably just manually instrument and trace through the C-tor implementation when I'm back from leave)