Fix calculation of client proof

[abiligiri]

• Pad g with 0s to be of size N
• Add variant for generating client secret that accepts password as data and without username

[adam-fowler]

• Pad g with 0s to be of size N

I don't see anything in the RFC about padding g with zeros before hashing it

• Add variant for generating client secret that accepts password as data and without username

This is a very odd addition in that passwords are strings. How come you need a password as Data?

EDIT: I take this back. The specification around padding is quite vague. The Mozilla vectors expect padding, so I'll take that as expected. Your PR doesn't cover all the cases. I have another PR that does this #14. If you could verify that is works for you that'd be awesome

[adam-fowler]

Now that #14 is merged this is no longer needed