Add trace_id and span_id fields to rich JSON log lines

[aapzu]

Is your feature request related to a problem? Please describe.

When running the agent with --write-job-logs-to-stdout and --log-format=json, the agent emits structured ("rich") log lines that already include contextual fields like org, pipeline, build_id, job_id, etc. However, when OpenTelemetry tracing is enabled, there is no way to correlate these log lines back to an active trace or span. This makes it difficult to cross-reference logs with distributed traces in observability backends (e.g. Datadog, Honeycomb, Jaeger).

Describe the solution you'd like

Add trace_id and span_id fields to the jobLogger constructed in NewJobRunner (agent/job_runner.go, around lines :301–319), populated from the active OpenTelemetry span on the context at the time the logger is created. For example something like:

if tp := r.conf.Job.TraceParent; tp != "" {
    // W3C traceparent format: {version}-{trace_id}-{span_id}-{flags}
    if parts := strings.SplitN(tp, "-", 4); len(parts) == 4 {
        log = log.WithFields(
            logger.StringField("trace_id", parts[1]),
            logger.StringField("span_id",  parts[2]),
        )
    }
}

This would allow operators to filter structured logs by trace_id in their log aggregation tool and jump directly to the matching trace.

Describe alternatives you've considered

I haven't come up with any actual alternatives. There's already the BUILDKITE_TRACING_TRACEPARENT environment variable, but that exists inside the agent, not outside it. therefore parsing the logs and injecting the fields from outside the agent isn't possible.

Additional context