The fix involves moving the function leb128Decode over to parser. To me
it makes more sense for the function to belong in that module so I think
of it as a positive change. However I do not think that returning two
values is really necessary. I think a proper solution would be either
to parse the code or wrap the stream so we can count how many bytes are
readed. Therefore we could use std.leb.readUleb128 which should be less
error-prone.
