Add Intel Xeon support (SPR+) for RAG [dev]

[tpawlows]

Same as #126, but for dev branch.

Please merge it after rh-ai-quickstart/ai-architecture-charts#137 is merged. - merged ✅

• This PR extends the quickstart with an option to deploy RAG on Intel Xeon for balanced price/performance.
  • It is similar to the CPU deployment but uses a container image with an optimized vLLM for Xeon that leverages AVX512 and AMX instruction extensions for improved inference (opea/vllm-cpu-ubi:v0.12.0-ubi9).
• By default, requires OpenShift cluster with at least one worker node that is using SPR or newer generation CPU with more than 16vCPU and 64GiB of memory, e.g. m8i.8xlarge (32vCPU128GiB SPR), m8i.8xlarge ( 32vCPU128GiB GNR)
• Validated models:
  • meta-llama/Llama-3.2-3B-Instruct
  • meta-llama/Llama-3.1-8B-Instruct
• Added example Xeon configuration to deploy/helm/rag/values.yaml
• Updated README.md
• To deploy use flag DEVICE=xeon:

  # Xeon deployment
  make install NAMESPACE=llama-stack-rag LLM=llama-3-2-3b-instruct SAFETY=llama-guard-3-8b DEVICE=xeon

• Succesfully deployed and tested with Add Xeon deployment support to llm-service Helm chart ai-architecture-charts#137