The column index POC kept track of how many "gaps" were skipped when scanning with filters. This information was used when a column with a value not matching its filter was found. Normally the zig-zag scan would iterate this column until a matching value is found. Instead, we would only iterate the column when the fraction of passes where a gap was skipped was low (< 10%). Otherwise the algorithm would continue as normal, assuming that it was likely another gap would be skipped because of missing values in certain columns.

We should consider adding something similar to this for the SBE column scan, as it would help with queries that contain some fields which are often missing and others which are not.