Fix escaped, escaped_transform, satisfy, one_of and none_of interpreting [u8] as UTF-8 (#1679)

[micolous]

Fix a number of issues relating to nom decoding [u8] buffers as UTF-8 when it shouldn't (#1679):

• escaped, escape_transform now accept control_char: impl AsChar, rather than char.

  This allows the functions to be used with a u8 or b'', which is useful for parsing text-like files that contain binary or non-UTF-8 data (like Lua).

• escaped, escaped_transform, satisfy, one_of and none_of now iterate over individual bytes when using a [u8] input, rather than attempting to interpret bytes as UTF-8 sequences.

I've added many tests demonstrating some edge cases of these functions. Many existing tests incorrectly used str for [u8] inputs in some places, which can lead to some unexpected behaviour when handling binary data.

This will probably break API compatibility for a parser that takes a [u8] buffer as inputs and assumes everything is decoded as UTF-8. I'd argue this is incorrect usage anyway – those should be using str.

There's probably other parts of nom that assume [u8] is encoded as UTF-8, but searching for these is hard.