Optimize equihash implementation

[svyatonik]

This is the first part of #64 . I've eliminated some allocations + extra loops && measured performance. The test was running verify_equihash_solution_works for 1000 times. Here are some results:

1. old version of parity-zcash Equihash: 0.02655900600075256 secs / 1000 runs;
2. librustzcash Equihash: 0.01449104799394263 secs / 1000 runs;
3. new version of parity-zcash Equihash: 0.010416991004603915 secs / 1000 runs.

So the new implementation is ~61% faster than the previous one && ~28% faster than the librustzcash. q.e.d.

[wkarshat]

@svyatonik great improvements.
Would help to spec model and speed of the CPU and RAM used in testing, to calibrate expectations.