feature: handle MSK events

[joeyzhao2018]

https://datadoghq.atlassian.net/browse/SLES-2739

In Kafka's wire protocol (KIP-82), header values are always byte[]. Every Kafka client library enforces this:

Tracer	Injection code	Mechanism
dd-trace-java	headers.add(key, value.getBytes(UTF_8))	String.getBytes() → byte[]
dd-trace-go	Value: []byte(val)	Go type conversion → []byte
dd-trace-dotnet	_headers.Add(name, Encoding.UTF8.GetBytes(value))	UTF8.GetBytes() → byte[]

All three tracers accept string trace context values from the propagation layer, convert to UTF-8 bytes at the carrier adapter boundary, and hand byte[] to the Kafka client.
This isn't a quirk of Java's getBytes() — it's the only way Kafka headers work.

What MSK Lambda does

When MSK triggers a Lambda, AWS serializes the Kafka record to JSON. Since header values are byte[] on the wire, AWS represents them as JSON arrays of unsigned byte values —
regardless of which language produced the message:

"headers": [{"x-datadog-trace-id": [51, 54, 57, ...]}]

What's the difference between the msk_event.json and the newly added msk_event_with_headers.json here?

msk_event.json represents a standard MSK trigger where the producer didn't attach any Kafka headers — i.e. no Datadog tracer was running on the producer side (or it's a non-instrumented producer like a raw Kafka client, a Kinesis Firehose delivery stream, or a schema-registry message). In those cases Lambda still delivers the event but with "headers": []. It's also the format you get when testing MSK triggers manually in the AWS console, which doesn't inject headers.
The msk_event_with_headers.json case corresponds to a producer instrumented with a Datadog tracer (e.g. a Java/Dotnet/Go service) that injects trace context as Kafka headers — AWS then serializes those header values as byte arrays when delivering the event to Lambda.