Standardize dtypes

[ashsaki]

• Use explicit fixed width types such as uint32_t instead of unsigned int.
• We are capping number of qubits to MAX_UINT16. However, group inds representing qubits use a mix of unsigned int and uint16_t. Standardize everything related max number of supported qubits to uint16_t.
• Use smallest data type possible everywhere.

[nonhermitian]

Yeah the number of qubits, and thus term indices can be uint16_t. The number of terms is currently size_t and I am not sure if we want to change that or not, i.e is 4 billion terms enough? The indexing of terms within groups, e.g. the group_ptrs necessarily needed to follow the size of the terms.

[ashsaki]

num_groups can stay size_t or uint64_t to be safe. Shortening qubit indices to uint16_t may help the cache hit rate.

[nonhermitian]

Item 2 is taken care of in #32