MILESTONE: CLIENT-CHOSEN PIXEL DATA IS NOW VERIFIED INSIDE THE XIE SERVER,
DISTINCT FOR SOLID-WHITE AND SOLID-BLACK, WITH NO DEBUG POKES.

THE RECIPE (all client requests, no pokes, reproduces):
  1. InitSession                 minor 1,  2 words  [major=2 | minor=3<<16]
  2. CreatePhoto (photomap)      minor 8, 12 words  [type=2, id, w=64, h=64,
                                                     0,0, w6=1, 0,0, 0x100, 0]
  3. SetTransport                minor 20, 35 words [photo_id, 0, 0, plane_mask=1,
                                 zeros x6, zeros x24]  with
                                 word2 = (mode=3 << 8) | (component_organization=1 << 16)
                                 pixel_stride[0] = 1, scanline_stride[0] = 8
  4. PutTile                     minor 19, 8 + 128 words
                                 [photomap, x=0, y=0, w=64, h=64, plane_mask=1, final=TRUE]
                                 + 128 words of tile data

RESULT -- the server's own element buffer:
  white tile:  element+0x28 = 008F8B90
               buffer[0..48] = ffffffff... (48 bytes, distinct = {255})
  black tile:  element+0x28 = 008F8B90
               buffer[0..48] = 00000000... (48 bytes, distinct = {0})

  QueryResource on the photomap: w=64 h=64 comp_COUNT=1 constrained=1

  And at export, the render driver reports a real source for the first time:
      RD_ENTRY
      RD_ENAB  flag=1
      RD_ALLOC rp=008FD090
      RD_SRCBASE a5=003FD590 e28=002909B0
      RP       a11=008FD090 src_base610=002909B0      <-- previously always 0
  (src_base = *(element+0x28), per the store at 0x1B238.)

WHY IT WAS ALWAYS ZERO BEFORE: transport `component_organization` selects a
handoff branch; org=4 (XieK_BitByPlane) makes the handoff clear element+0x28
(0x6B8D6), while **org=1 preserves it**.  With org=1 the source survives to the
renderer.  Earlier traces showed src_base "-" only because the RP breakpoint was
not in that run's set.

STILL NOT DISPLAYED: Export (minor 21) returns err2, and framebuffer captures of
the 64x64 window are unchanged (white vs black differ by 1186 px = the Terminal
Manager's own figures and the cursor).  The remaining failure is on the export
side, AFTER the renderer has a valid populated source.
