Treat UTF-16 strings in binary VDF as little-endian

[smcv]

Integers in binary VDF are already treated as little-endian (least significant byte first) regardless of CPU architecture, but the 16-bit units in UTF-16 didn't get the same treatment. This led to a test failure on big-endian machines.

Resolves: https://github.com/ValvePython/vdf/issues/33

[smcv]

~~I still need to test this on a big-endian machine, I'll mark as non-draft after that.~~

[smcv]

The CI failures are expected, see #56 for fixes for those.

[smcv]

Verified to pass tests on a Debian-developer-accessible s390x (big-endian) machine.