Update self-hosted models generations and suggestions to use v3 code suggestions endpoint

What does this MR do and why?

Removes AiGatewaySelfHostedMessages so that requests made from air-gapped customers can use v3 generations with added context.

How to set up and validate locally

1. Start AIGW with this MR - feat: adds model_params to v3/v4 completions (gitlab-org/modelops/applied-ml/code-suggestions/ai-assist!2184 - merged)
2. Configure you gdk to use self-hosted models for suggestions
3. Trigger a completion after a comment or in a small file - You should see /v3/completion being called. Sometimes /v2/completion is called because completion uses v2 prompts in all cases.

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

Related to #515582 (closed)