Update docs for str::as_bytes_mut.

Update docs for str::as_bytes_mut.

* Add "Safety" section describing UTF-8 invariant.

* Remove mention of `from_utf8_mut`.  It is not necessary to call
  a function to convert the byte slice back to a string slice.  The
  original string becomes accessible again after the byte slice is
  no longer used (as shown in the example code).