should really work on gage tests before more re-architecting

for third derivatives:
 * have to update gageScl3PFilter_t and change arguments to not
   itemize filter weights or derivative answers, involving
   sclfilter.c, scl3pfilterbody.c,
 * add gageSclJerk  (full third derivative)
 * considering third-deriv-abort directory:
   NO: it does make sense to talk about second derivatives with a
   2-support kernel: the small kernel still has second derivatives,
   even if its constant zero.  With tent, for example, the mixed
   partials will still be non-zero

recovering the pre-ISBI12-optim work
 * ctx.c: removal of gageIv3Fill
 * iv3Gage.c: new home for new gageIv3Fill, which includes some
   type-specific shortcuts
 * stack.c: use of new stackIv3FillBody.c
 * filter.c: some nrrdKernelHermiteFlag-specific smarts (line 386)
 * gage.h: gageContext->hermiteNonZeroPvlIndex (line 649)

figure out how to handle derivative renormalizating (for scale) at scale=0

add flags for explicitly specifying gmax, hmax, vrange

6pack filtering, adapt vector stuff

change type of iv3Fill to take (int xi,int yi,int zi) instead of void
*here; one step towards supporting bricking.

test: two pervolumes with different sized volumes
test: NRWS doing its job

retest comparison of upsampling by probing and by unu resample

make the curvature direction code (scl.c) more robust
