Wider data types

[alxvth]

• Use size_t for container size
• Use std::uint64_t and std::int64_t instead of unsigned int and int to prevent overflows
• Add template parameters typename unsigned_integer and typename integer to most classes in order to use 32 bit ints if desired
• Harmonize knn search in HierarchicalSNE and HDJointProbabilityGenerator
• Increase version since
  • consumers of this library now need to specify some types like hdi::dr::GradientDescentTSNETexture<float> instead of hdi::dr::GradientDescentTSNETexture, see update example
  • the cpp version changes to 17 from 14

Fixes #28.