Make ngm work on computers with more than 256 cores

[hermannschwaerzlerUIBK]

This is a proposed fix for #34:

• Use std::thread::hardware_concurrency() to find the available cores and use this number as num_devices parameter when calling clCreateSubDevices() instead of 256.
• Add "-std=c++11" option to compiler. Needed for the above when using older gcc (like e.g. 4.8).

As I don't have access to Apple computers I am not sure if the second change might have to be done differently (outside the IF (NOT APPLE) clause?).