Byte-swap integers to support big endian architectures

[onitake]

Fixes: #47

libbgcode always processes binary data in a byte-wise fashion. This breaks when the host architecture is not little endian, because binary Gcode requires data to be in little endian byte order.
This PR separates byte-wise and integer I/O and checksum functions, so the latter can be byte-swapped before they are processed. All occurrences of integer data access that I could find have been replaced with calls to the new functions.

The PR also removes the redundant write_to_file and read_from_file functions in core.cpp and binarize.cpp and moves them to a single implementation in core_impl.hpp.

The patched code has been tested on amd64 (little endian) and sparc64 (big endian). All unit tests pass.

Possible further improvements:

• Replace store_integer_le and load_integer with boost::endian, which might make the code easier to understand and more optimal in some cases, although modern compilers can figure out what the functions do and replace them with optimized built-ins in some cases.