Cython headers for Google Sparsehash

This library provides Cython .pxd files for Google's Sparsehash library. Both dense_hash_map and sparse_hash_map are implemented. They are useful only when used directly for Cython.

In case you need to use sparsehash from Python, there is an incomplete wrapper of sparse_hash_map in class SparseHashMap which will require setting the C++ template parameters manually. The current version maps uint32_t to uint16_t as that is what I originally needed.

Refer to sparse_hash_map documentation for available methods and to test_internal.pyx for example usage.

TODO:

• Add a custom hash function to SparseHashMap
• Benchmark cdef get versus cpdef get with a custom hash function
• Figure out a way to instantiate SparseHashMap without setting template parameters manually
• Figure out a way to provide __sizeof__ in SparseHashMap
• Consider migrating to serialize/unserialize of sparse_hash_map. The docs say that it is currently the preferred way to do I/O. However, the only advantage I see is that it is also implemented by dense_hash_map, unlike {read,write}_{metadata,nopointer_data}.