Optimise code

Some of the data processing code is quite slow when dealing with large data sets.
Could potentially be sped up with the use of dtplyr.

No file, test, entry point, dataset, or benchmark is named. Start by locating the data-processing paths in the R package and profiling them on a representative large dataset; confirm with the maintainers which workload and performance target define done, then verify that results remain unchanged.

Written by the indexing model from the issue text.