Memory leak in the latency tracker.

[Stebalien]

No requests are running but we appear to be leaking a bunch of these.

leak.zip

[dirkmc]

Possible cause: It looks like we start recording latency when RecordPeerRequests() is called and we remove the request when we receive

• a response (by way of peerData.AdjustLatency())
• a broadcastTimeoutMessage
• a peerTimeoutMessage

But if we receive a peerTimeoutMessage after the want was cancelled it looks like we don't call recordResponse() so peerData.AdjustLatency() doesn't get called and the latency request tracker doesn't get cleaned up.

[dirkmc]

@Stebalien could you please give repro steps to produce the memory leak profile so I can make sure it's fixing the same issue you saw?