Filter unavailable OpenAI limits

[shiny-code-bot]

Summary

• Filter ChatGPT-backed Codex additional model limits against the live /backend-api/models availability response, while failing open if availability is unavailable.
• Keep active buckets when either the display limit name or metered feature matches an available model identifier.
• Make stale/failure widget states red and clickable, but only route to reconnect when provider failure evidence exists; age-only stale data prompts refresh instead.
• Add an app attention/reconnect view that matches provider reports to configured account IDs where possible.
• Document the model-availability display filter behavior.

Review notes

• codex review --uncommitted found multi-account reconnect matching, stale false-reauth, and metered-feature filter risks; all were addressed before commit.
• code exec --sandbox read-only ran a second release-blocking review pass over the uncommitted diff.

Validation

• swift build
• swift test --filter 'WidgetSnapshot|Codex|codexConnector'
• swift test
• scripts/commit-gate.sh
• uv run /Users/cbusillo/Developer/codex-skills/jetbrains-inspection/scripts/jb-inspect.py closeout --repo "$PWD" --scope changed_files (clean, 0 problems)
• scripts/context-panel-runtime-baseline.sh install --launch (baseline=OK app=/Applications/Context Panel.app)
• scripts/context-panel-runtime-baseline.sh check (baseline=OK app=/Applications/Context Panel.app)