Compute dasher pay from order event logs

You are given event records for delivery orders handled by a single dasher. Each record is a dictionary with these keys: - order_id - event_type, one of {CREATED, ACCEPTED, PICKED_UP, DELIVERED, CANCELED} - timestamp, an ISO 8601 datetime string that may include a timezone offset or end with Z Records may be unsorted. For each order, pay starts at the first ACCEPTED event and ends at the first later DELIVERED or CANCELED event. An order contributes pay only if it has a valid ACCEPTED -> terminal pair after sorting its events chronologically. Apply these rules: 1. Ignore CREATED and PICKED_UP events for pay calculations. 2. If an order never has ACCEPTED followed later by DELIVERED or CANCELED, that order contributes 0. 3. If multiple events for the same order have the exact same timestamp, process them in this order: CREATED, ACCEPTED, PICKED_UP, DELIVERED, CANCELED. 4. If a timestamp has no timezone, treat it as UTC. 5. For each order, active minutes = floor((end_time - accept_time) / 60 seconds). 6. Total pay = sum(active_minutes * rate_per_minute) across all orders. 7. Return the total pay as a float rounded to 2 decimal places. Implement a function that parses the datetimes correctly, reconstructs each order's relevant state transitions, and returns the total pay.