ENH: CPU implementation of `WarpFieldBSpline` class

Builds on !21, which must be merged first.

This MR adjusts the WarpFieldBSpline class so that it can be compiled to run on either GPU or CPU. All this required was to move the existing routine for calculating the jacobian determinant into a standalone function (moved to src/WarpFieldBSplineUtilsGPU.cu), and to add a new equivalent CPU-based routine (in src/WarpFieldBSplineUtilsCPU.cpp).

This MR also adds a new --debug_dir command-line option which, when passed along with --debug causes debugging information to be saved to a user-specified directory, instead of <pwd>/debug/.