"Char" C type are going to be 16 bits (unicode)
The most widely used encoding of unicode, UTF-8, is based on regular
bytes (it's a variable length encoding backwards compatible with
Ascii). That means both legacy character strings and unicode strings
will use byte operations. Protocol networks, compression formats will
also remain byte-based. And so on ;)
Gimp use 16 bit per channel
I think that's false, that's even why Film Gimp exists (movie guys
need more than 8 bits per channel and Gimp can't satisfy their needs).
However even if Gimp can do 16 bits per channel, most images will
still only require 8 bits per channel, leading to more compact
representations and faster calculations (and less memory transfers
;-)).
Also if you think SIMD in stuff like GIMP, operations with 8-bit
per channel will be twice faster than with 16-bit channel (because
your registers have a fixed size and you can stuff twice more bytes
than 16-bit integers into a register).
Half the bandwith consumption but many more operations to avoid