merge from visdev to ML

[Luka <luka@MIT.EDU>]

i've merged stable changes from my visdev branch to mainline.

this does *not* yet include the proximity culling hack.  that
and further vis development will go in and continue to happen
on branch.

this *does* include a huge load of other good changes i've done
since december, summarized below.  first a log of my merge
commands, then a summary of contents the merge delivers to ML.

merge commands...

in existing visdev branch sandbox:
  cvs tag visdev_luka_03-1218_mergeto_ml_04-0323_tag

in new dir for merge to ML:
  cvs co walkthru
  cvs tag visdev_luka_03-1218_to_ml_04-0323_premerge_tag
  cvs up \
    -j visdev_luka_03-1218_branchtag \
    -j visdev_luka_03-1218_mergeto_ml_04-0323_tag
  cvs commit
  cvs tag visdev_luka_03-1218_to_ml_04-0323_postmerge_tag

in english:
  this tags head on visdev as the point i consider stable so far,
  tags head of mainline with the premerge tag,
  applies the delta from start of branch to that point onto ML,
  commits the new head rev and tags it with the postmerge tag,
  so the resulting pre-to-post tag change is the merge.
  (all modulo anyone else's concurrent activity, but there was none.)

merge contents...

libs (incl binaries, scripts, and configs) were previously mixed
up between the checked in tree and the external system deps, and
specified such that certain system configs could lead to incompat
code getting quietly connected either at compile or link or run time.
great badness resulted.  fixed.  current state is as follows:
  libtiff is still checked in as binary for some-platforms-only.
    (todo: fix this, prob making it an external devel lib dep too.)
  tcl/tk (and blt if anything uses it) are fully removed from the tree.
    the system (or your environment) must provide something resembling
    tcl/tk version 8.3 for walkthru to work.  all our systems do this.
  all this is tested heavily on irix, and somewhat on linux.
    (todo: check linux.  and make dynamic linking it work right, since
    at least redhat was getting increasingly hostile toward static.)

a bunch of tapestry code was cleaned up (excised).  because doing the
whole job is insane, but doing some was required, i did this:
  removed tapestry from main dm/vis render path, and cauterized the
    ends that this left loose elsewhere.  (so render path now is
    "good" and rest of system is "stable".  but the remains of tapestry
    are crippled and should not be activated.  remove as convenient.)
  removed all code from util/tapestrygen dir. (this one thing was easy.)

sync'd up a bunch of duplicated and independently modified code that
should have been merged long ago, or never even split.
among notable results are 3 similar files that matter a lot:
  skel/skelmain.cpp     : main skeleton file for use as example only.
  firewalk/firewalk.cpp : firewalk's main file.
  furn/firewalk.cpp     : furn's main file.
that is, these 3 files are similar but still substantially different in
functionality (and can't be trivially compacted). 
  note that the job is not done (but this was the most important part).
there are still a mess of other files which are dup'd between
firewalk/furn and should get resolved/collapsed whenever a test build
of furn finally succeeds.

did a full review of ndb module.

various visualization bugfixes.  things relating to -DGL.  wireframe mode.

unlimited observer motion in UIMoveObserver.  (just a start to the big
changes.  this part just frees the UI controls.)

removed old-style semaphore comments... new system is fully in place now.