Allow AIGW timeout setting on GitLab Dedicated instances

What does this MR do and why?

Allows GitLab Dedicated admins to configure the AI Gateway request timeout setting, which was previously only available on self-managed instances.

The ai_gateway_timeout_seconds setting was gated behind manage_self_hosted_models_settings, which is blocked on Dedicated instances (via self_hosted_models_allowed → self_hosted_enabled_for_dedicated?). This caused Dedicated customers to be stuck with the default timeout, leading to A1000 errors when LLM responses exceeded it.

Changes:

• Add a new update_ai_gateway_timeout permission, enabled for admins. Unlike manage_self_hosted_models_settings, this is not blocked on Dedicated instances.
• Update the DuoSettingsUpdate GraphQL mutation to check the new permission for ai_gateway_timeout_seconds (in addition to manage_self_hosted_models_settings).
• Expose can_manage_aigw_timeout from AiConfigurationPresenter to the frontend.
• Move the timeout form in the Vue component to render independently of the self-hosted models block, controlled by canManageAigwTimeout.

References

• #603860 (closed)
• https://gitlab.com/gitlab-com/request-for-help/-/work_items/4918 (internal)

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist.

Closes #603860 (closed)