Uses arrow::open_dataset() with predicate pushdown and column projection for lightning-fast querying and minimal memory overhead on daily hydroclimatic records.
Adaptive routing based on CSV size and opening-cost heuristic, avoiding unnecessary scans for medium-sized selections.
Added start_date and end_date parameters to load_pe_timeseries() for temporal slicing at scan time.
Added a lightweight self-contained sample dataset in inst/extdata/sample_camels_pe (~120 KB) enabling autonomous execution of examples and tests without requiring external internet downloads.