Add e2e test for EPP metrics scraping

[shotarok]

What would you like to be added

Add end-to-end (e2e) tests to verify that EPP metrics can be properly scraped from the metrics endpoint at port 9090.

Why is this needed

As discussed in PR #980, we need comprehensive e2e tests that include test cases for scraping EPP metrics with assertions for the expected metrics. While we have integration tests that verify metrics functionality, we currently lack e2e tests that validate the full metrics scraping workflow in a real deployment scenario.

The EPP currently exposes the following metrics:

• inference_model_request_total
• inference_model_request_error_total
• inference_model_request_duration_seconds
• normalized_time_per_output_token_seconds
• inference_model_request_sizes
• inference_model_response_sizes
• inference_model_input_tokens
• inference_model_output_tokens
• inference_model_running_requests
• inference_pool_average_kv_cache_utilization
• inference_pool_average_queue_size
• inference_pool_per_pod_queue_size
• inference_pool_ready_pods
• inference_extension_info

Implementation Details

The e2e test should:

1. Deploy an EPP configuration
2. Send inference requests to generate metrics
3. Scrape the metrics endpoint at port 9090
4. Assert that expected metrics are present with appropriate values
5. Verify metrics behavior under different scenarios (successful requests, error conditions, etc.)

Related Issues

Resolves discussion in PR #980: #980 (comment)

[shotarok]

Although there is no e2e test, there is hermetic_test for EPP's metrics. It might be enough for the verification.

[delavet]

Is this issue duplicated with #915 ?

[shotarok]

@delavet That's true! Thank you for letting me know. I'll close this issue.