Fix writing half code points to Android logger

[middagj]

This fixes if a Unicode code point is incorrectly split when writing it to the Android logger. Both split_at (split_at_checked) and floor_char_boundary use is_utf8_char_boundary. I added a method that combines both functions. Also floor_char_boundary is relatively new (1.91).

Second commit is reducing the fully unsafe function where already separate unsafe blocks where introduced. I don't know if the unsafe fn was intended as a marker, but it also makes it possible to unintentionally introduce more unsafe calls in the function.