[Obs AI Assistant] Use preconfigured elser inference endpoint

[sorenlouv]

Problem
The Observability AI Assistant currently has a setup step for creating an inference endpoint obs_ai_assistant_kb_inference. When created this will download .elser_model_2. This is required to use the knowledge base and adds complexity to the codebase.

Solution
8.17 ships with pre-configured inference endpoints for elser and e5 models (elastic/elasticsearch#116931). We should replace our own endpoint obs_ai_assistant_kb_inference with .elser-2-elasticsearch:

kibana/x-pack/plugins/observability_solution/observability_ai_assistant/server/service/kb_component_template.ts

Line 64 in f6ac2cf

inference_id: AI_ASSISTANT_KB_INFERENCE_ID,

Current setup UI

This screenshot shows the setup UI that we can get rid of if we use the preconfigured inference endpoint

[elasticmachine]

Pinging @elastic/obs-ai-assistant (Team:Obs AI Assistant)

[sorenlouv]

This has been postponed because the preconfigured endpoints don't allow us to adjust min_number_of_allocations. As soon as that is possible or we have a multi-tenant inference service we can pick this up again. In the meantime we'll be using custom inference endpoints.