feat: add support for TPUs on llm-d reference architecture

[syeda-anjum]

Overview

This PR adds support for Gemma-4 on TPUs with llm-d, including Kustomize templates and a new README guide.

Key Changes

• Added docs/platforms/gke/base/use-cases/inference-ref-arch/llmd/llmd-vllm-with-hf-model-tpu.md.
• Added Kustomize manifests for v6e-gemma-4-26b-a4b,v6e-gemma-4-31b and 'v6e-qwen3-32b' in online-inference-tpu/llmd/vllm.

Impact of Change

Enables users to deploy Gemma-4 and Qwen-3 on TPUs using llm-d and provides documentation for it.

References

• https://llm-d.ai/docs/guide/Installation/optimized-baseline
• https://github.com/llm-d/llm-d/tree/main/guides/optimized-baseline