Databricks model services silently ignore thinking effort

**Bug:** Goose accepts thinking-effort changes for Claude model services, but Off/Low/Medium/High/Max produce identical MLflow requests with no reasoning configuration.

**Reproduce:** Select a fully qualified Claude model under databricks_v2, change thinking effort, and inspect outgoing requests.

**Expected:** Apply supported effort settings, or disable the control when unsupported.

Look for the Databricks model service integration code, likely in a module like `src/services/databricks_v2.rs`. Find where thinking-effort settings are processed for Claude models and where MLflow requests are constructed. The bug is that the effort setting is ignored in the request. Check if the Databricks API supports a reasoning configuration field. The fix is to map the effort levels to the correct API parameter or disable the UI control if unsupported.

Rédigé par le modèle d'indexation à partir du texte de l'issue.