Speed optimizations for poseidon2_goldilocks_plonky2 and field.

[Mychecksdead]

Speed + memory optimizations.

Speed:

• Removed branchHints as they aren't working.
• Manually inlined MulAccF.
• Some changes to externalLayer function (reordering of calculations, manual inlining).

Memory:

• Created HashNToMNoPadBytesOptimized. This new function assumes a 32-byte output array as parameter.

• Similarly updated Sum function.

• Directly used bits.LittleEndian.PutUint64 function to avoid make() inside ToLittleEndianBytes.

• Added tests.

• Simple changes in uint128 for simplicity.