Reliable Shopify order processing with webhook monitoring

“**Problem** A webhook delivery is one HTTP attempt (internal/webhook/webhook.go deliverSingleWebhook). A receiver that restarts, answers 503 or 429, or is unreachable for a moment loses the event for good, and nothing records it. Receivers that must not miss events (for example a bot that answers customers, or an SLA and paging integration) end up polling the conversations API to find what they missed. **Proposal** - Retry network errors and 408, 425, 429 and 5xx responses; 2xx and other 4xx end the delivery. - `webhook.max_attempts` (first attempt included) and `webhook.retry_backoff` (first delay, doubling, capped at 10 minutes). A missing `max_attempts` keeps today's behaviour, one attempt. - Retries are scheduled with a timer and re-enter the delivery queue, so a slow or dead receiver does not hold a worker. Before each retry the webhook is read again, so a disabled or edited webhook stops getting it. - Every attempt sends the same body and signature plus three headers: `X-Libredesk-Event`, `X-Libredesk-Delivery-Id` (one UUID per delivery to one webhook, the same on every attempt) and `X-Libredesk-Delivery-Attempt`. Receivers can drop duplicates by the ID. - Test deliveries from the UI are not retried. - Pending retries live in memory, like the queue today, and are lost on restart; a durable outbox with a delivery log in the UI could follow if you want it. **Compatibility:** additive headers; behaviour changes only when `max_attempts` is set (the sample config sets 4).”