windsock: measure consume latency

[rukai]

Updates to latest windsock, making use of the new consumer latency functionality.
To calculate the consumer latency we:

• serialize the current SystemTime in the first 12 bytes of the produced record payload
• take those bytes from the consumed record payload and turn them back into a SystemTime then call .elapsed() to determine the amount of time passed.

This matches the behaviour of rdkafka_performance: https://github.com/confluentinc/librdkafka/blob/master/examples/rdkafka_performance.c

I did not use a serialization library for this as I need to ensure they fit exactly into 12 bytes so we know what our minimum size of payload needs to be.

As a result of the new minimum of 12 bytes, the existing 1B bench case was turned into a 12B bench case.

[codspeed-hq]

CodSpeed Performance Report

Merging #1753 will not alter performance

_{Comparing rukai:windsock_consume_latency (8415fe5) with main (4853f84)}

Summary

✅ 39 untouched benchmarks