Performance improvements, memory leak and misc other fixes and drop binary size by ~4%

[cbnilrem]

Biggest change is the introduction of base classes to dedupe shared code. Drops binary size by ~4%, but the tradeoff is slightly reduced type safety at the lowermost interop layer. As that layer is now mostly hidden by helper layers I feel the tradeoff is worth the improvement. Next biggest change is a cache of function table pointers to avoid unnecessary marshal calls.

[johnkord]

Signed off, some good refactoring in here