Improve bitstring initialization efficiency

[zaqqwerty]

When sampling state circuits, it would be more efficient to replace symbols in a tiling of bitstring gates before concatenating with the model, than it is to scan for the bit symbols in an already concatenated circuit.

[zaqqwerty]

Since this is the bottleneck for deep circuits, need to time three ways of sampling circuits:

1. Current implementation
2. tile bitstring circuits alone then concatenate to concrete model
3. tile full symbolic circuit, resolve bitstring and model params simultaneously

If this is still a bottleneck, are there further changes that could help?