[Gtkradiant] IMap branch merged back into trunk / experimental version of map module

[Timothee Besset]

I merged back the IMap branch into the trunk after finishing a very first 
working version of the map module. There is still a lot of issues and 
things to deal with, but it passes the test of loading a map and saving it 
back.

What are the most significant changes:
- the extraction of the parsing/saving code into a module was a rough 
process. Basically the Map_LoadFile and Map_SaveFile functions have been 
moved out and compiled in a module. This required to add a whole lot of API 
calls from modules into the core (see ibrush.h and ientity.h). It was the 
opportunity to experiment with some macros to maintain the same syntax for 
common stuff across the core and the modules. For instance you would use 
Entity_Alloc in the core or in a module the same way, but in a module it 
would be expanded to g_EntityTable.m_pfnEntity_Alloc

What's left to do:
- some outdated code still lives in the core (old functions used for 
parsing the map and writing it back)
it can be removed now
- debugging and stabilization. The main issue is memory alloc/free on 
win32. See http://www.qeradiant.com/faq/fom-serve/cache/179.html for 
details or ask here. Basically, when MSVC breaks in _dbg_free it means you 
are trying to free a memory block that was not allocated in the same module 
(i.e. .dll or .exe)

Long term todo:
- the way we fill in the function tables in the core is awkward in some 
places. The whole scheme to share function tables across modules and fill 
them intelligently will be abstracted in a seperate library.
- a new map module can be written to handle .xmap format, the existing map 
module will have to be extended to support Q1 and Q2 native file formats 
etc. (this work can probably start right now)

TTimo