chore: mark BitVec.flattenList noncomputable

The structural definition is only a specification; the divide-and-conquer
flattenListFast (installed via @[csimp]) is the intended runtime implementation.
Marking flattenList noncomputable removes the slow compiled version entirely, so
the csimp'd implementation is guaranteed to be the only one ever used (and
flattenList drops off the csimpCore audit list of computable @[csimp] targets).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>