Add: mca control for cpu load balancing (and don't report Gflops figures for cpus we can't determine)

[abouteiller]

Add an mca control that disable cpu-gpu load balancing (disabled by default). This can be overriden in multiple ways:

• setting mca param will enable cpu load balancing globally
• providing an evaluate function for the CPU type can set the selected device
• data_advise_on_device can be set to a CPU, etc.

Change reporting of flops for devices for which it cannot be determined, it will now print 0.0 if we can't figure it out, and use default time_estimate using 1.0 as a baseline in that case.