feat: Databricks model registry refresh mechanism

## What does this PR do?
Allow the Databricks model list to be refreshed at runtime so deprecated models don't silently 404.

## Why
Flagged during the v0.6.0 release review (CTO persona). v0.6.0 hardcodes 11 Databricks foundation models. When Databricks deprecates one (e.g. `databricks-gpt-5-mini`), users hit silent 404s with no clear recovery path.

## Proposed
Two options:
1. `databricks list-models` CLI — hit `/api/2.0/serving-endpoints` and diff against registered models, warn on drift.
2. Config override — let users declare custom Databricks models in `altimate.json` (same pattern as custom OpenAI-compatible models).

Probably both. Document in the Databricks providers section.

## Deferred from
v0.6.0 release review. Filed at tag time per no-follow-up-PRs release policy.

## Type of change
- [x] New feature