Small files can skew upload time statistics

In a small swarm, if a peer uploads a lot of small chunks, it will get unprortionally good upload time stats. Maybe change the statistics to bitrate instead (and calculate timeouts from there) to avoid this?

Start by locating the upload-time statistics and timeout calculation entry points in the repository, then trace how small chunks affect peer scoring in a small swarm. Compare the current time-based measurements with a bitrate-based approach and define tests that confirm small-chunk-heavy uploads no longer receive disproportionate statistics while timeout behavior remains valid.

Written by the indexing model from the issue text.