Follow-up from "Fix hypervisor sub-leaf handling"

The following discussion from !460 (merged) should be addressed:

• @smcv started a discussion:

  guint is only 32 bits, and GHashTable keys are pointer-sized (therefore 32-bit when compiled for i386), so a hash table { guint => SrtCpuidData } cannot provide correct mock values for something that wants 64-bit keys.

  I see two ways this could go:

  • short-term: declare that items with a non-zero subleaf just can't use this mocking mechanism, use something like G_UNLIKELY (mock_cpuid != NULL && subleaf == 0), and go back to using GUINT_TO_POINTER (leaf) as the hash table key;
  • long-term: use a data structure with larger keys and a custom hash/equals function, like a hash table { SubleafPair* => SrtCpuidData* } where SubleafPair is a struct { unsigned leaf, unsigned subleaf }