Changes for adding gemma model config

[vavarshn]

Summary

This MR updates the AI Tier SAIA model deployment config to use Gemma 4 as the primary hosted SAIA v2 model while keeping GPT-OSS 20B available for supporting flows.
Changes include:

• Replace the GptOss120b Ray Serve application with Gemma431bIt (gemma4_31b_it).
• Keep GptOss20b deployed for field descriptions, conversation titles, and metadata-description paths used by SAIA service.
• Update SAIA feature replica defaults to scale Gemma431bIt and GptOss20b.
• Update the Ray builder config test to expect Gemma431bIt + GptOss20b as the text generation apps.
• Update k0s quickstart model artifact documentation from gpt-oss-120b to gemma-4-31b-it.

Why

SAIA v2 hosted model selection now defaults to Gemma 4 when use_gpt_oss_120b is false or missing. This operator change makes the corresponding Gemma 4 Ray endpoint available in AI Tier and removes the GPT-OSS 120B app from the local deployment config.
GPT-OSS 20B is intentionally retained because saia-service still uses it for auxiliary CMP flows such as field descriptions, title generation, and metadata descriptions.