Consider running ai-gateway E2E tests against CNG pipeline

With CNG pipeline being a faster alternative to the Omnibus build, and with the Omnibus build not being run in MRs by default. It could be beneficial to run the ai-gateway-related tests in the CNG pipeline. Currently the ai-gateway job runs tests tagged with :ai_gateway against the latest ai-gateway in mock response mode.

ai-gateway has a helm chart: https://gitlab.com/gitlab-org/charts/ai-gateway-helm-chart

Note: i notice at the above link ⚠️ This helm chart is meant for self-hosted models users only!. Although we would be using it in with AIGW_MOCK_MODEL_RESPONSES=true so I am not sure if it really matters what model we are using.