And other minor code cleanup.

Without the flex/bison generated files, it's very hard to build the code
on Windows.

Contributors must simply remember to keep these files up-to-date if the .l
or .y files change.

Only tested with Visual C++ 2008.

Update ruby bindings for libconfig

- Fixed incorrect buffer switching.
- Fixed crash in scanctx_cleanup() when file list was empty.
- Fixed config_init() to properly set the default include function.
- Fixed memory leak when freeing file list.

- Make use of scientific notiation configurable (%f vs. %g) via new option.
- Set default float precision to 6 (was 2).
- Fix clipping of negative values when casting to unsigned int in C++ library.
- Fix bug in formatting of floating point values (removal of trailing 0's)
- Split tab width and float precision into two separate fields.

hook functions are now used so the app can supply its own logic.

Various other cleanup.

Rake::GemPackageTask has been deprecated in favour of Gem::PackageTask.

As of Ruby 2.2 RFLOAT is private, use RFLOAT_VALUE to obtain the value.

Currently only has complete Win32 implementation. POSIX implementation
is incomplete.

Robustness & bug fixes for tab_width handling

 * Reset default float_precision to 2 (from leftover DBL_DIG)

 * bugfix (missing mask) that multiplied indentation by 3840 (0x0F00)

 * Robustness: convert get|set_tab_width() from macro to function
   Also: const correctness

Locating libconfig from CMake-based projects.

Tested. Looks good.

Implement float_precision parameter (Closes #31)

Added the ability to specify the amount of decimal digits (precision)
to be used when writing out (file or stream) a configuration.
Includes C & C++ versions.

As per zhaopingsun's suggestion

Merge fix for issue #41 from random85 ; Documentation and fixes by me