Include a null terminator for byte[].

[dfish3r]

See #44.

[dfish3r]

I'm not certain what should be done if the input is already null terminated.
I added a check so that the terminator isn't added if it already exists.
I think the spec could be read either way.

[serac]

I think adding iff not exists is perfectly reasonable, especially in the absence of evidence one way or the other. In the end, though, it's all about client support, so if there's an important implementation that always adds a null terminator we could justify doing that for the sake of interoperability.