Handle Typeform API rate limits and back off correctly

Documented steps

1. Auth: Authorization: Bearer {access_token}; the 2 req/s budget is scoped to the account owning the token.
2. Documented limit: 2 requests per second per Typeform account for the Create API and Responses API — throttle clients to <=2 req/s (token bucket or sleep-based).
3. The limit is per account (the account owning the token), not per token — multiple workers sharing an account share the budget.
4. Webhooks and Embed SDKs are not subject to this client-request limit (webhooks are Typeform calling you).
5. Typeform error responses always include a `code` field; `description`, `field`, and `in` may also appear for validation errors — branch retry logic on `code` (retry 5xx/503, don't retry 4xx validation errors).
6. The exact status code returned on exceeding 2 req/s is not explicitly documented — handle 429 and 5xx defensively with exponential backoff rather than assuming one specific code.
7. Don't conflate this with webhook DELIVERY retries (Typeform calling your endpoint: 30s timeout, tiered retry schedule) — separate system, separate rules.
8. Official docs: https://www.typeform.com/developers/get-started/ ; https://www.typeform.com/developers/troubleshooting/