Cutoffs exist #
The source asks for a smooth even cutoff, valued in [0,1], supported in (-1/2, 1/2) and
identically one on |x| ≤ 1/2 - delta. That is exactly Mathlib's ContDiffBump centred at the
origin with inner radius 1/2 - delta and outer radius 1/2: the bump is radial, so evenness
comes free from ContDiffBump.neg, and the two radius conditions are what 0 < delta < 1/4
supplies.