Solidity transcripts optimization, improvements

[philippecamacho]

• Append point/Field Elems: serialize into bytes, then append
• little-endian vs big-endian in solidity
• overhead in ser/de serialize
• fields elems as u256, basic word in solidity
  • Benchmark gas usage

https://gitlab.com/translucence/cap-on-ethereum/cape/-/issues/122

[alxiong]

another potential source to be optimized:

    function fromLeBytesModOrder(bytes memory leBytes) internal pure returns (uint256 ret) {
        // TODO (optimization): Can likely be gas optimized by copying the first 31 bytes directly.
        for (uint256 i = 0; i < leBytes.length; i++) {
            ret = mulmod(ret, 256, R_MOD);
            ret = addmod(ret, uint256(uint8(leBytes[leBytes.length - 1 - i])), R_MOD);
        }
    }

(since Transcript.sol depends on BN254.fromLeBytesModOrder())