Fix issue when parsing large CPU Masks (+32C) systems

[benjojo]

Previously when using 32Thread+ systems, the libary would misparse the cores beyond the 32T threashold, causing bad behaviour.

This has now been fixed, with tests (and the parsing being moved to it's own testable sub-function)

Tested on a 12C*HT*2S, 32C*HT, and 32C*HT*2S system's output on modern linux kernels