force_bytes should handle the common case first like force_text

[Aymeric Augustin]

The patch I'm about to commit makes force_bytes twice as fast in the common case — and twice as slow when it's called with a memoryview, but we don't use them that often in Django.

Before:

% python -m timeit -s 'from django.utils.encoding import force_bytes' 'force_bytes(b"")' ~/Documents/dev/django
1000000 loops, best of 3: 0.705 usec per loop

After:

% python -m timeit -s 'from django.utils.encoding import force_bytes' 'force_bytes(b"")' ~/Documents/dev/django
1000000 loops, best of 3: 0.376 usec per loop

[Aymeric Augustin <aymeric.augustin@…>]

In b80be68e088f56f5b2ab5371b92a478a14937e8b:

Fixed #21052 -- Small performance optimization.