Currently, all API logic follows different code paths depending on whether the node is a leader or a follower. This should be changed so that decisions are based on the availability of the required components (e.g., the ingest or stable table) rather than the current role.

The availability checks for tables, and potentially other state that currently depends on the node role, should be performed before entering the business logic. The business logic itself should be moved into separate functions that operate independently of the current role.

This will better prepare us for async step-up and elegant step-down, while also making layered cursor unit testing simpler to implement.