Fix Auth Resnponse packet when cleartext is used

[methane]

Fixes #884

Description

NUL shouldn't be after string[n] auth-response. It breaks following string[NUL] database.

https://dev.mysql.com/doc/internals/en/connection-phase-packets.html#packet-Protocol::HandshakeResponse41

So n should contain trailing NUL byte.

I confirmed cleartext is used in sha256_password. I can't confirm cleartext password.
Especially, there are no way to enable fast cleartext password in MySQL 5.7 and 8.0.
(default-authenticate-plugin can't be "cleartext_password".)

Checklist

• Code compiles correctly
• Created tests which fail without the change (if possible)
• All tests passing
• Extended the README / documentation, if necessary
• Added myself / the copyright holder to the AUTHORS file

[methane]

One problem is what should be returned for empty password: empty string or one byte NUL.

Whan I tested with sha256_password (which returns cleartext when secure connection), one byte NUL is required. I suppose cleartext password plugin is same.

[julienschmidt]

Why did you remove the addNUL?
It's there to avoid an allocation for a return value which is just copied again afterwards anyway.

[methane]

Why did you remove the addNUL?
It's there to avoid an allocation for a return value which is just copied again afterwards anyway.

Because I thought it's there to add NUL after string[n]. And appending it after n bytes was the bug.

I think both of []bytes(mc.cfg.Passwd) and append([]bytes(mc.cfg.Passwd), 0) are allocate once.

Anyway, code around addNUL looks ugly, and this code is used only for authentication, not query.
So performance difference should be ignorable.

[jpeirson]

I can confirm that the fixes in 369b5d6 fix the issue I was having.