Ensure we're never overlong

[klnusbaum]

From @Cwillmore on discord

This is neat! one thing you should be aware of is that it's illegal for a utf8 string to encode a unicode codepoint in more bytes than the least possible. for example it's only legal to encode character 0x41 = 0b01000001 'A' as 01000001 (and not 11000001 10000001). in utf8 land this is called an "overlong form"
https://kevinboone.me/overlong.html?i=1
i think it's possible for your cipher to end up emitting an overlong form, e.g. if you encode two \0 chars in a row

Let's write some tests to ensure we're never overlong