Design structure for bottom-K customers by revenue

You are building an in-memory data structure to track customers’ **current revenue** and answer “top K” queries, where **top K means the K smallest revenues**. Implement a class/data structure that supports: 1. `upsert(customerId, revenue)`: Insert a new customer with the given `revenue`, or update an existing customer to this new `revenue`. 2. `bottomK(k) -> List[customerId]`: Return the IDs of the **k customers with the smallest current revenue**. Requirements / clarifications: - `customerId` is unique. - Revenues can repeat across customers. - If `k` is larger than the number of customers, return all customers. - If multiple customers have the same revenue, return them in increasing `customerId` order (or state and apply any deterministic tie-break). - Aim for efficient time complexity; be ready to discuss trade-offs for **read-heavy** workloads (many `bottomK`) vs **write-heavy** workloads (many `upsert`).