Refactors Data.checksum() increasing performance 5 times.
Tests on MacBook Pro 2018 (2,3 GHz Quad-Core Intel Core i5, RAM LPDDR3) show for data of random 1_000_000 bytes:
* original implementation 0.586 sec
* proposed implementation 0.126 sec
* also tried `UInt16(bytes.lazy.map(UInt32.init).reduce(UInt32(0),+) % 65535)` which takes 0.417 sec