Secure LLM display name now shown in the chat attribution pill

The friendly display name administrators set for the Secure LLM provider — for example, “Our EU GPT-4o” — now appears under each assistant reply served by it, replacing the raw model identifier. The raw model id moves to the pill’s hover tooltip for transparency.

The label is captured as a snapshot at reply time and persisted with the message, so renaming the provider later does not retroactively rewrite history, and the friendly name survives page reloads and shared-conversation views. Replies without a configured label continue to show the raw model id.

A small schema addition is applied automatically on startup. No action required.