[Clas_offline] mc recon problems fixed, gsim updated

[Paul Mattione]

Thanks to all of the help I received (especially from Franz Klein and  
Eugene Pasyuk), my monte carlo reconstruction is now working!  For  
photon runs after g11, you must do the following:

  - check out and compile the new version of gsim
  - in your tcl file for user_ana, include "set st_type 1" and "set  
photon_trig_type 1"

The gsim package needs to be recompiled because the geometry code for  
the long photon target (40cm) and new ST (STTYPE 1) wasn't correct, in  
particular for other target positions than -10cm, because some  
position information was hard-coded in gsim (updated gsim/geom_tg.F,  
gsim/geom_phtg8.F). In addition, there were missing offsets for the  
hit banks of the start counter, causing the STN TDC value to be wrong  
for tracks with vertex <-20cm (gsim/sets_stnew.F).

Also, the C++ code for new ST reconstruction (set st_type 2 in  
user_ana tcl file) doesn't work for monte carlo data unless you copy  
all database geometry values for your experiment into run #1, along  
with tweaking st2tof as necessary to line things up.

Thanks again everybody for all of the help!

  - Paul