The big-endian logic here was simply incorrect. Luckily, it was also
overcomplicated; after calling `Value.writeToPackedMemory`, there is a
method on `std.math.big.int.Mutable` which just does the correct
endianness load for us.