Lantern: more accurate H2 simulation

Right now the server response time isn't included per request and we really need 1 RTT per initiator

i.e. an asset is only "free" once we've incurred at least 1 RTT since all of their dependencies have been met

Complicated but would help accuracy on h2 simulation.

Start by locating the Lantern H2 simulation entry point and reviewing how dependency completion and request timing are currently modeled. Trace where per-request server response time and RTT assumptions would be represented. Done means the simulation applies one RTT per initiator before an asset becomes free and has coverage for the updated timing behavior.

Written by the indexing model from the issue text.