Refactors kafkajs tests so that instrumentation errors are easier to see

[codefromthecrypt]

This refactors the tests for kafkajs to use v2 format and fixes some lower hanging problems

• docker containers didn't cleanup which made debugging mysterious
• setup was being used same way as client-server, which is incorrect
  • remote service should be the broker, not the guessed name of a server
  • producer and consumer should never share span ID (join) especially because this library does retries for failed message processors!
• incorrectly read Buffer as a string which broke propagation
• incorrectly set remote service name where service name goes
• mislabled span names differently than the calling functions

Notable concerns left out of scope are:

• consumer span is being labeled on a processor function, which is a modeling glitch. Consumer should represent the time spent getting the message from kafka, which is quite short.
• docker setup is ok now, but still could be better possibly if managed by mocha somehow instead of shell script

See #384

[codefromthecrypt]

this still needs work to do as there's a bunch of timeout problems in both kafka via docker, and also the sequencing of promises in general. I suspect we should be looking at testcontainers-node.. cc @bsideup in case you know an existing setup

[codefromthecrypt]

we might consider moving this logic into traceId, if it is common for headers to be of type Buffer, though I think Buffer might be node.js only and so far this is the only user of it.

[codefromthecrypt]

ps the timeouts are not so much a problem of tests not running (they do) it is more about the obfuscation of issues. It isn't as high priority to manage test differently, just it is hard to maintain especially the test that looks for errors to occur.

[codefromthecrypt]

in local dev, tests would fail because the pre-existing docker containers weren't deleted. the result would be strange as it would appear parsing was busted because the same headers are always read... this was due to replaying whatever the first test message was. now this is fixed, but really we have to switch this to something else!

[bsideup]

@adriancole one can run the tests with GraalVM and enjoy the Java version of Testcontainers :D
https://medium.com/graalvm/using-testcontainers-from-a-node-js-application-3aa2273bf3bb

[codefromthecrypt]

OK I rewrote the description. I'm done with this work because there are a lot more modules to refactor and the most important things here are addressed now.