This commit fixes the following problems converting to/from Python integers
and ffi types:
- integers of 8 and 16 bits not working on big endian
- integers of 64 bits not working on 32 bits architectures
- unsigned returns were converted to signed Python integers
Fixes issue #7269.