Parse bytes rather than chars

[NathanBaulch]

Very handy functions, just what I needed!

I was looking over your parse function and couldn't help but wonder if processing the ulid as bytes rather than chars might improve performance. Turns out it does, by a factor of about 2x in my tests!

-- parse a million sequential ULIDs
select count(parse_ulid_mine(concat('0', i))) from generate_series(1e24, 1e24 - 1 + 1e6) as i;
-- mine: 23.619s
select count(parse_ulid_yours(concat('0', i))) from generate_series(1e24, 1e24 - 1 + 1e6) as i;
-- yours: 49.648s

FWIW, I'm running Pg 16.4 in Docker in WSL on Windows on a 7 year old notebook.