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

We look for the setting using its path and extract the setting name from the
path to get its index in parent for removal.

This adds "Removing a setting with its path" feature.
This changes the specification, the documentation has to be updated.

The test doesn't behave as documented. The wrong element is removed.

Octal ints