PIPELINE PATH CRASH DIAGNOSED -- IT NEEDS A FULL IDS CONTEXT, NOT JUST A FLO.

gdb backtraces (decclient argv[4]=2):

  attempt 1:  XieCrop(pm,0) -> XieTapFlo(crop,1)
              SIGSEGV in XieTapFlo ()            [crop had no flo link]

  attempt 2:  XieCreatePhoto(XieK_Photoflo) -> XieBindMapToFlo(flo, pm)
              -> XieCrop(flo,0) -> XieTapFlo(...)
              SIGSEGV in XieBindMapToFlo ()       [before any element work]
              (the photoflo itself was created: printed photoflo=555555580770)

So the DEC library's element/bind entry points fault when driven from a
hand-built client, even in the order DEC's IDS code implies.  Reading
ids_build_xie_pipe.c shows why: the IDS layer maintains a full xiectx/phoctx
context (XieIdc templates, destination resolution via _XieGetDst, taps created
on an already-populated chain) -- this is not a short sequence of public calls,
it is the DECimage application library driving them.

CONCLUSION: this avenue cannot be exercised without building the IDS layer
(ids_build_xie_pipe.c + ids_render_xie.c + ids_xieimage_ddif.c, ~4,700 lines with
unsatisfied external dependencies), which is a separate integration project.

STATE
  * NO client-chosen XIE image has ever been displayed;
  * pixmap export path runs clean but paints nothing (engine does no transfer);
  * pipeline/bind path faults inside libXie;
  * VRAM byte-identical for white and black (0 differing bytes / 128 KiB);
  * accelerator model opt-in (VT1200_ACCEL=1); default unchanged;
  * xshow (core X) is the only viewer that has shown an image.
