If we encounter an error at measurement n while attempting to stage ordered writes to the bucket catalog, we populate the errors vector for all measurements after and including n. This is done here. However, this is a change from the existing behavior where we would only populate the errors vector with the error for the nth measurement (here, here).

This will cause existing tests like `bulk_write_timeseries_basic.js` to fail.

We only populate the nth measurement and all subsequent measurements when encountering a staging error in the ordered path if we fail while trying to acquire the collection (here, here)