RDataFrame seems too conservative about spawning new threads

Yes, my code involves a kinematics fit on top of a toy MC to do benchmark, ( pdk2025/record_pdk.cxx at kf_augmented · karuboniru/pdk2025 · GitHub ) and each fit ( pdk2025/src/kf.cxx at kf_augmented · karuboniru/pdk2025 · GitHub ) would take ~ ms of time to finish.