tokenizer: propagate trace context on vLLM render requests

[mschulist]

Wrap the render HTTP transport with otelhttp.NewTransport so that outbound /render requests carry the W3C traceparent header. This links the tokenizer-to-vLLM hop into the same trace span as the EPP request, matching the existing pattern used by the sidecar proxy transport.

[github-actions]

🚨 Unsigned commits detected! Please sign your commits.

For instructions on how to set up GPG/SSH signing and verify your commits, please see GitHub Documentation.

[Copilot]

Pull request overview

This PR instruments the tokenizer’s vLLM /render HTTP client transport with OpenTelemetry’s otelhttp transport wrapper so outbound render requests propagate W3C trace context (traceparent), aligning tokenizer→vLLM tracing behavior with the existing sidecar proxy transport pattern.

Changes:

• Wrap the vLLM render http.Client transport with otelhttp.NewTransport(...) so trace context is injected on outbound requests.
• Add a unit test asserting that a traceparent header is present and carries the expected trace ID.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
pkg/epp/framework/plugins/requestcontrol/dataproducer/tokenizer/vllm_http.go	Wraps the render transport with otelhttp.NewTransport to propagate trace context on outbound /render calls.
pkg/epp/framework/plugins/requestcontrol/dataproducer/tokenizer/vllm_http_test.go	Adds a test validating traceparent injection for render requests.

Suppressed comments (1)

pkg/epp/framework/plugins/requestcontrol/dataproducer/tokenizer/vllm_http_test.go:329

• After adding a synchronization channel for the handler, wait for it before asserting on gotTraceparent so the test has a defined happens-before relationship.

	_, _, err = r.Render(ctx, fwkrh.PayloadMap{"prompt": "hello"})
	require.NoError(t, err)

	if gotTraceparent == "" {
		t.Fatal("expected traceparent header to be injected into outbound render request, got none")