Make sure ScanScheduler works well with multiple object stores

Currently, the ScanScheduler is designed to work with a single object store. When multiple object stores are involved, this can lead to performance issues and just bad defaults for various configs because the configs assumes 1 object store connection.

In the multi-base implementation, the quick workaround I did was to create one scheduler per object store, but I remember when I discussed with @westonpace during that time, the intended design was that there should only be a single scheduler, and it should instead reference multiple object stores.

This is combing back again in the external blob case @Xuanwo did in blob v2, as external blobs can have different prefixes thus using different object stores, so we need a way to manage the scan scheduler with multiple object stores.

Creating this issue to track this.

Start by tracing ScanScheduler and the multi-base implementation described in the issue, then compare the one-scheduler-per-object-store workaround with the external blob v2 case. Done means a single scheduler can manage multiple object stores and no longer assumes one object-store connection when choosing configuration defaults.