OutputWrapper's design led to easy-to-introduce string mismatch

issues, which could cause crashes or data corruption. This is because it

was meant to be used with either a Unicode or byte string buffer, and

calls to .write() failed to handle all cases of mismatches.
This change fixes this all through a combination of type hints and

smarter string conversions.
OutputWrapper is now a generic, available in OutputWrapper[bytes] or

OutputWrapper[str] form. The form used dictates the type of upstream

buffer this can work with, and the types of string values allowd when

writing.
To protect against bad calls, we now use force_bytes() or

force_unicode(), depending on the type. This happens automatically for

all writes. This also allows us to simplify all the internal logic.
Since OutputWrapper.write() defaults to using a newline as the end

value, and we can't hard-code a string type here, we now use a wrapping

object that symbolizes a newline. The prior hard-coded default is likely

why the write() logic was the way it was before.
Unit tests have been added for the class testing both bytes and str

variations.

Unit tests pass.
Tested command output from a few different commands, utilizing both

Unicode and byte streams. Didn't encounter any issues.