Add latency calculation tests & logs

[cesher]

Added witness latency calculation tests. Specifically the following scenarios are now being tested:

• the request and the response have 8 ms of latency
• the request and the response have 0 ms of latency
• the request and the response overlap by -1 ms of latency
• the request and the response fully overlap each other have 0 ms of latency
• the request and the response are in the wrong order have -4 ms of latency
• the request and the response timestamps are the zero value have 0 ms of latency
• a stream with two interleaved request and response each with 3 ms of latency
• two pipelined requests and responses have 0 ms of latency

Added 2 validations in the stream processing layer:

• if the first or last packet timestamps are the default zero value, override to time.Now().
• if the last packet timestamp is before the first packet timestamp, override last packet timestamp to first packet timestamp.

^These are being logged and printed as part of the PrintWarnings report at the end of a session.

Added 3 validations in the backend collector layer:

• if the collector matches 2 requests or 2 responses together (due to HTTP/1.1 pipelining and matching on ACKs), log it and leave latency default value of 0.0.
• if the collector sees timestamp zero values, log it and leave latency default value of 0.0
• if the collector sees negative latency calculation, log and report it.

[mgritter]

Looks OK -- do you plan on enabling debugging in beta when you deploy this? (Or do we already have debug log level set?)

[mgritter]

This is good but we should figure out some more K8s-friendly way of showing our counters, for example including them in telemetry or dumping them out periodically to the log.

[cesher]

I'll do this in a follow up PR, I have to first update the schema in the shared libs repo before I can push these values through the telemetry worker.