chore: Added end to end tests

[shmuelk]

This PR adds an initial set of end to end tests for the llm-d-inference-scheduler. The test causes the llm-d-inference-scheduler to be built. The tests are run under a Kubernetes Kind cluster that is setup and torn down as part of the tests. The llm-d-inference-sim is used for the vLLM servers in all of the tests.

These tests include:

1. A simple EPP configuration with a single vLLM simulator.
2. A setup that uses Prefill/Decode disambiguation with one prefill instance and four decode instances.
3. A setup that uses the precise prefix scorer exploiting KV events with a single vLLM simulator.

There is a set of utility functions that possibly can be upstreamed for common use in tests both in the llm-d-inference-scheduler and in the upstream Inference Gateway. This will investigated at a later time.

[elevran]

Ref #209

[elevran]

/approve
/lgtm

[elevran]

/approve
/lgtm