Add vLLM+HPA support to ChatQnA Helm chart

[eero-t]

Description

Add vLLM + HPA support for ChatQnA Helm charts.

Similarly to how it's already done in Agent component, there's tgi.enabled and vllm.enabled flags for selecting which LLM will be used.

Note: ChatQnA does not yet support using vLLM for embedding & reranking [1], so HF TEI is still used for those.

[1] opea-project/GenAIExamples#1237

Issues

Fixes #608 partially.
Fixes #631.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds new functionality)

New dependencies

n/a

Tests

Manual testing on top of main HEAD / v1.1 images.

[eero-t]

Setting as draft. I have tested that DocSum with Gaudi vLLM works, and that ChatQnA Helm chart can be installed, but due to v1.1 image pulls currently taking so long in my test node, I haven't been able test ChatQnA with Gaudi vLLM properly yet.

vLLM CPU version testing would also be needed before merging this (I'm hoping somebody else here could check at least DocSum with CPU vLLM).

[eero-t]

CI issues:

• LLM-uservice: openai.NotFoundError: Error code: 404 - {'object': 'error', 'message': 'The model meta-llama/Meta-Llama-3-8B-Instruct does not exist.', 'type': 'NotFoundError', 'param': None, 'code': 404}
  • Bug already in the CI/repo, pre-existing Helm chart refers to model not present in CI => I can fix it
• DocSum: 100.83.111.229:5000/opea/llm-docsum-vllm:latest: not found
  • Bug in OPEA image creation, that image is missing from DockerHub & CI => somebody else needs to fix that

[eero-t]

This overlaps partly with #403.

[eero-t]

Added HPA support for ChatQnA / vLLM.

[eero-t]

This bug for llm-uservice, gaudi, ci-vllm-gaudi-values, common is weird:

openai.NotFoundError: Error code: 404 - {'object': 'error', 'message': 'The model `meta-llama/Meta-Llama-3-8B-Instruct` does not exist.', 'type': 'NotFoundError', 'param': None, 'code': 404}

Because neither that CI file, nor values.yaml specifies that model for llm-uservice:

$ git  grep -l '^[^#]*Meta-Llama-3-8B-Instruct'
helm-charts/common/llm-uservice/ci-faqgen-values.yaml
helm-charts/common/llm-uservice/variant_faqgen-values.yaml
helm-charts/faqgen/README.md
helm-charts/faqgen/values.yaml

So CI either invents that model from thin air, or is using wrong CI file (faqgen instead of vllm-gaudi one)!

[eero-t]

Changes:

• Cherry-picked changes from Adapt to latest vllm changes #632
• Dropped vLLM support for DocSum as required image is missing
• Rebased to latest main

[eero-t]

Rebased to main, and dropped fix for model ID [1].

[1] it's specific to DocSum & FaqGen specific llm-uservice variants. Text generation llm-uservice variant uses different env var name from those (which IMHO is bug, but for code in another repo). I'll add workaround for it in separate PR for DocSum vLLM suport.