Experimental SDE raster support for 4.4

[Howard Butler]

All,

Thanks to help from Pirmin and Frank, I have committed some changes to the
SDE raster stuff to allow the user to specify the important bits that
they'll need to actually be able to display rasters using PROCESSING
directives.

To use SDE rasters in 4.4, a user will have to manually define
USE_SDERASTER as well as the normal USE_SDE symbols.

The SDE wiki page has some info on the bottom that describes how I was able
to get a raster to display.
<http://mapserver.gis.umn.edu/cgi-bin/wiki.pl?MapServerSDE>.

ArcCrap 9 does some really funky things when loading a image through the
GUI tools.  By default, it seems to define a seven band image, with rgb
being in bands 4,5,6.  Copying one of these images that has already been
loaded into SDE into a new table with the GUI tools makes a normal three
band image, with rgb being bands 1, 2, and 3 like you would expect. This is
what was causing me problems in my test cases.

By being able to define the important bits with the PROCESSING directives,
I was able to get things working quite well.  I only tested with ArcSDE
9.0, but both my windows and linux test setups worked.

We still have some issues that will need to be resolved regarding SDE
rasters, and I think it should be listed as experimental until these
changes are made.  Different versions of SDE (8.0, 8.1, 8.2, 8.3, 9.0) seem
to have different capabilities regarding raster support (do we make config
files sensitive to this? yuck).  Although I am not sure, SDE 3.x doesn't
seem to have raster support.  I am doubtful that SDE for Coverages will
work with this either.  If anyone has information to contradict or confirm
my suspicions, please let me know.

Longer term, I am looking at moving the SDE raster stuff into GDAL, to get
all of the capabilites that Frank outlined in a previous email.  It would
be nice to get extra utility out of the code as well as eliminate possible
duplication of anyone's efforts with regard to raster data. Maybe we should
look at moving the vector side of into OGR as well.  Both would be a
substantial effort.  Anyone who is interested in helping do this now that
we have a good start, please contact me and/or Frank.

Howard