[matthias-kleiner]

New overload of TPCFastSpaceChargeCorrectionHelper::createFromLocalCorrection taking the knot positions in y and z as vectors, so a correction from a continuous function can have dense knots where it changes fast (e.g. at the sector edges). The knots are rounded to an integer grid with 1/10 of the smallest knot distance as step, as in createFromTrackResiduals; the first and the last knot are put at the ends of the grid (row y edges; central electrode and readout in z). The function is sampled at 6 points per knot interval over the whole grid.

The existing uniform version is unchanged. It samples y only between the outermost pad centres, so the outer half pad of each row is extrapolated by the spline. Test with a B = 0 cosmics correction map (spline - map, rms in dx / dy / dz):

• 10 x 10 uniform knots: new overload 61 / 42 / 43 um, existing function 60 / 49 / 42 um; within 6 % of the row width of the sector edges dy 41 um (1.7 mm max) vs 218 um (3.9 mm max).
• 20 x 10 uniform knots: new overload 59 / 38 / 39 um; existing function 0.5 / 2.1 / 1.0 mm, up to 16 cm on the pad range and 36 cm at the sector edges.