Helper class for saving measurement result distributions from multiple shots

[sethrj]

The #17 PR has a "buffer manager" which seems actually to be a class for storing measured distributions from multiple shots. With Qsim, every invocation of the entry point is a single shot, and we can run the execution engine in a loop to measure these and report them. (This is in contrast to XACC which implicitly performs multiple shots and can only run deterministic [single LLVM block] code.)

I think we want a standalone class that can let us accumulate final measurement states across multiple runs. A vector of qubit indices should map like a bitset, and the user interface (which can provide string outputs) should be the only thing that needs an "endianness" field. It should either be an array (with 2^N values for N qubits) or unordered map (if we expect that a only a tiny amount of the combinatorial measured results are possible).

In #17 the buffer class lives inside the qsim directory but instead we should have it in the main qiree directory.

[sethrj]

ResultDistribution I think would be a better name; "Buffer" is an implementation detail of XACC.