fix(pricing): add Fable 5.1 rates and Grok bot aliases

[robinebers]

TL;DR

Price Claude Fable 5.1 and alias Cursor's grok-bot-* slugs to Grok 4.6 so those unknown-model warnings stop eating spend.

What was happening

• Cursor usage listed unpriced models: claude-fable-5-1-thinking-{low,medium,high} and grok-bot-{automation,cua,default}.
• The existing Fable 5 alias only matches claude-fable-5, so the new 5.1 thinking slugs never resolved.
• Neither bundled LiteLLM nor models.dev snapshot carries Fable 5.1 yet.
• The Grok bot slugs had no alias at all.

What this changes

• Add a claude-fable-5.1 supplement entry from Cursor models & pricing and Anthropic's pricing page:
  • input $10
  • cache write $12.50
  • cache read $0.25 (0.025x; 75% below Fable 5's $1)
  • output $50
• Alias claude-fable-5-1 / claude-fable-5.1 plus thinking, effort, [1m], and Cursor Router labels (Fable 5.1 (Auto …)) to that entry. Fable 5 stays on its own catalog rates.
• Alias grok-bot-automation, grok-bot-cua, and grok-bot-default to grok-4.6 ($2 / $2 / $0.50 / $6).
• Bump updated_at so installed apps pick this up from the feed.

Heads-up

• Left other Cursor-page diffs alone for a later pass: Grok 4.5 Fast output is now $18 on Cursor's table (we still have $12); GPT-5.6 Sol is on promo $4/$20 there (we keep official $5/$30); Sonnet 5 is $2/$10 on Cursor (catalogs still $3/$15).
• grok-bot-* looks like Cursor-internal routing, not a published model. Priced as Grok 4.6 on purpose.

Tests

• swift test --filter "ModelPricing|PricingBundledResource": 57 tests passed.
• Supplement JSON / regex / multiplier validation passed.

Made with Cursor