0. General code clean-up, including messages, and the like.

0. General code clean-up, including messages, and the like.
1. Changed the name of the class `AliFMDMultStrip' to
   `AliFMDRecPoint' as there's no longer the parallel class
   `AliFMDMultRegion', and the class name was confusing.
2. I've overloaded `AliDetector::LoadPoints' to draw hits using the
   class `AliFMDPoints', which uses a `TMarker3Dox' instead of
   `TPolyMarker3D'.
3. `AliFMDGeometry' gets the transformations from the `TGeoManager'
   and stores these with each `AliFMDDetector'.
4. The geometry is modified, so that half rings/cones are separate
   objects, to allow for alignment of these.
5. Made the Poisson method of reconstruction a script class,
   derived from the utility class `AliFMDInput'.
6. Code gets calibration parameters from `AliFMDParameters' in a
   proper way (that is, by detector coordinates).
7. Made two member functions that translate from hardware address
   to detector coordinates, and back.
8. Added TODO file
9. Removed a lot of unused code.

What I see as missing:

1. Read calibration objects from CDB in AliFMDParameters.   To that
   end,  I intent to make a script that creates fake parameters in
   a file, and then try to read in those.
2. Make fake alignment objects via a script, and then see that
   these are applied properly, and that RecPoints end up at the
   right place (I've made the utility class AliFMDDisplay for that
   purpose among other things).
3. The proper trailer should be written to raw files.  However, the
   recent changes by Cvetan are not flexible enough for the FMD to
   use it.   I've sent Cvetan a mail and a proposal on how I'd do
   it.   Unless something like that is accepted into the RAW
   module, I think I'd have to go with something else.   Not an
   ideal situation, but I'm afraid that AliAltroBuffer just doesn't
   meet our requirements.
4. Implement calibration classes for pedestals and gains.
   However, these will probably only serve as examples and for
   testing, as the real programs will run on the LDC's, probably
   not using AliROOT.
5. Implement algorithms to go from survey points to AliAlignObj.
   However, that really requires that we know where the survey
   points are, which is currently not the case.