C++ cleanups, field-only merkle tree, LongsightF removal

[HarryR]

This makes the following changes:

• Significant C++ cleanups, using VariableT, ProtoboardT etc. throughout, including namespaces
• Removes LongsightF
• Moves Miximus to LongsightL-MP compression function
• Merkle tree path authenticator which uses field elements instead of bits

The proving key for the Miximus circuit is now down to ~550kb and runs very quickly, and components are tested in both Python and Solidity. However the following items remain:

• [ ] Solidity implementation of Merkle path authenticator
• [x] Python<->Miximus(C++) tests

[HarryR]

Interesting warnings from clang build:

/Users/travis/build/HarryR/ethsnarks/src/mod/miximus.cpp:77:15:warning:moving a temporary object prevents copy elision [-Wpessimizing-move]
        m_IVs(std::move(merkle_tree_IVs(in_pb)))
              ^
/Users/travis/build/HarryR/ethsnarks/src/mod/miximus.cpp:77:15:note:remove std::move call here
        m_IVs(std::move(merkle_tree_IVs(in_pb)))
              ^~~~~~~~~~                      ~
1 warning generated.

I had problems with the Miximus circuit during proving where it would access an invalid variable (e.g. offset in the hundreds of trillions range). I think I 'fixed' it using std::move and an rvalue.

However, another problem is that I turned some compiler warnings off in the GCC build which made the compilation output cleaner (all the unused variable warnings from libff).

But, the good news is the Python & C++ implementation of miximus work together well, so just need to cleanup these last warnings.

However, now the Mac build is erroring with the following message in the middle of the tests, without indicating which test it was running:

make: *** [cxx-tests] Abort trap: 6

I will have to double-check the Mac build locally.