Now parallelize the evaluation of the populations

Update installation file and add a new cmake build method. Type 'cmake .', then make and the installation process stay the same ( . install.sh {local})

-CUDA.tpl corrected bugs
-Improve the user information, by dumping proper population size to distribute to current GPGPU system.
-Resolve a bug in GPU evaluation. Now evaluation are really done in parallel on all the cards, (it was serialized before).
-Add instrumented tools in CEvolutionaryAlgorithm.cpp (INSTRUMENTED must be defined to use this feature).
     -If timing.csv file does not already exist, a description of each field is outputted first in the file.
-It is now possible to specify a range of GPU to be used for evaluation computation at execution time. This is done using --fstgpu and --lstgpu arguments, that specify used gpus as [fstgpu,lstgpu[.
-Installer improved for linux and MacosX

	Cuda code generation was bugged due to this problem.

	A lot of modification, mainly about the mode of every source files.
	I removed the "exec" bit of these files.

	Now to install easea :
		make
		sudo make install
	for the user :
		make vars
	for the developper :
		make dev_vars