Fix the result mapping for a customer charge

An assistant documented chargeCustomer as throwing on a declined card and rejecting amounts that are not positive. The code did neither. A decline came back as a plain object, and callers treated it as a success. Your team picked one contract and made every documented branch real.

An assistant wrote charge_customer_result(status, charge_id, decline_code, amount_cents), which maps the gateway's answer to that contract. It is in the editor. It has a bug. Find it and fix it.

Input and output:

• status: the gateway's status for the charge, like "succeeded" or "failed". It can be any string, including "".
• charge_id: the gateway's charge id, or "".
• decline_code: the gateway's reason for a decline, like "card_declined", or "".
• amount_cents: the amount charged, in cents.
• Return a list of two strings: the outcome ("ok", "declined" or "error"), then a detail.

Rules the tests check, in this order:

1. If amount_cents is 0 or less, return ["error", "invalid_amount"], whatever the gateway said.
2. If status is exactly "succeeded" (case matters), return ["ok", charge_id]. If charge_id is empty, return ["error", "missing_charge_id"] instead. decline_code is ignored here.
3. Every other status is a decline. Return ["declined", detail]. The detail is decline_code, or status when decline_code is empty, or "unknown" when both are empty.