Fix serialization / deserialization of char array.

[moriyoshi]

This patch fixes the following issues:

• Deserialization of char[] which contains bytes in range of 128 to 255.
  For example using unpack spec '>hh" on "\xff\xff" results in (-1, -1) and causes ValueError in bytes().
• Correct handleing of surrogate pairs.
  Just fixing the above doesn't solve all the problems. Applying.decode('utf16-be') on a resulting buffer of such two bytes that represent surrogate pairs ends up in UnicodeDecodeError.
• Debug logging of non-ASCII characters in range of U+1000 to U+FFFF
  I don't really think that it is a good idea to manipulate strings for debugging regardless of its being enabled or not.
• Proper Python 2 support.
  In py2 bytes is merely an alias of str, which doesn't construct a sequence of bytes from a iterator.
• Serialization of char[].
  It's not been implemented at all.

[coveralls]

Coverage decreased (-0.7%) to 84.357% when pulling c26690f on moriyoshi:moriyoshi/fix-type-char-unpack into 8472ed9 on tcalmant:master.

[coveralls]

Coverage decreased (-0.7%) to 84.357% when pulling c26690f on moriyoshi:moriyoshi/fix-type-char-unpack into 8472ed9 on tcalmant:master.

[coveralls]

Coverage decreased (-0.7%) to 84.317% when pulling 38c7b92 on moriyoshi:moriyoshi/fix-type-char-unpack into 8472ed9 on tcalmant:master.

[tcalmant]

Thanks for your contribution!
I'll make a release after some tests and completing the AUTHORS file.

[moriyoshi]

Thank you for looking into this!