feat: add configurable model shortcuts for Slack threads

[makeitfutureDev]

Change

Add gateway-wide model shortcuts in Settings → Access & security → Model shortcuts. Admins map names to an engine and a model from its current catalog. After selecting the bot mention in Slack, type :astra after Slack's inserted space, optionally followed by a task. This pins the resolved engine/model pair to the thread; a bare shortcut switches it without spawning an agent turn. Later edits to mappings leave already selected thread pins intact.

The shortcut honors runtime-change permissions, disabled engines and dedicated Codex-login restrictions. Runtime-card retries remove the original shortcut prefix from the task while retaining the runtime selected on the card.

Validation

• Static checks passed: 781 JavaScript files, no undeclared identifiers, no artifact secret writes, clean whitespace.
• Seven focused settings/pipeline tests passed. The pipeline runs actual Claude/Codex fixture processes for new turns, resumed turns and engine switches.
• One explicit Chromium test passed: create, save, reload, repoint engine/model, remove; no page errors.
• Full suite: 3,211 passed, 30 skipped, one existing README link-count assertion failed. README.md and test/readme.test.js are unchanged from the beta base; the recent public documentation link makes a count expected to be one equal two.

Live provider checks and private QA records remain pending and are documented as unexecuted in TEST-PLAN.md. This is a beta integration, not stable-release acceptance.