Improve error estimation on ridge points identified with KDEs

Currently, the default is to estimate error on the ridge points as the bandwidth of the KDE. There are also options to estimate the error as the full-width half-max (FWHM) of the KDE and via bootstrapping. The FWHM and bandwidth estimators result in an overestimation of the error bars. However, performing a reasonable number of bootstrap resamplings (on the order of 10^3) takes too long to run without HPC.

No file, test, or entry point is named. Locate the KDE error-estimation implementation and compare the bandwidth, FWHM, and bootstrap paths, then determine an approach that avoids the reported overestimation without requiring roughly 10^3 bootstrap resamplings on non-HPC systems.

Written by the indexing model from the issue text.