The fieldExtract struct escapes to the heap and the textChunks and numbers slices in fieldExtract.Extract are allocated every time, pulling these allocations into a reusable struct that can be passed to ParseDate would greatly speed up date parsing.

Jira issue: CRDB-21450