Implement suited unique set size (USS) for Linux

[lexover]

Implement suited unique set size (USS) for record memory activity which more sutable for Linux systems using argument --linux.

[astrofrog]

Thanks for the pull request!

This is a good idea but I think we can simplify this somewhat - basically you could just check to see if current_mem.uss raises an AttributeError or not - if not, then include it in the results, otherwise we don't. Then there is no need for a --linux flag and for the version checks for psutil. The check for current_mem.uss could be done in the first loop and could set a has_uss flag that could then be used in subsequent iterations.

Let me know if you don't have time to work on this, and I can push some commits to this pull request to show what I mean.

[lexover]

Thanks for the review!

The USS (Unique Set Size) works for Linux, OSX, and Windows platforms. If method memory_full_info () from psutil will be implemented for OSX or Windows the results of memory usage for this OS will be changed from RSS to USS without any warning. I think it is not good. At the moment in the code of psutil for Linux OS implemented and documented the memory_full_info(), and for Windows and OSX implemented but not documented memory_addrspace_info() methods which calculates USS. Perhaps it makes sense to implement for Linux use USS by default (--linux flag will be removed) or using the --uss param and later implement these function as the option for all appropriate OS if it is necessary (Unfortunatly I cant test this methods on OSX and Windows. I can implements this, but it's needs to be tested on real OS).