- 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.

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

Add @include_dir directive: debian-style ("conf.d") config processing

Apply stylistic fixes, following Mark's advice: homogeneous style
with the rest of the code

 - Misc documentation fixes

 - Document the include_dir directive
 - Robustness: filter out non-files (where scandir supports it)
 - Add myself to the credits

  Quite some refactoring of the includefile stack and filename processing logic

  Adds some extra fields to scan_context, but does not change the API
  "symbols" file should hide all functions called scanctx_* anyway

Configuration files for CMake are provided to simplify
locating libconfig from within project that use CMake
as their build system.
Separate configuration files are provided for C and C++
builds.

In a CMakeLists.txt file, use the following commands to
locate libconfig:

- To locate the library, use either
    find_package(libconfig)
  or
    find_package(libconfig++)

- The include directories are available as variables
    LIBCONFIG_INCLUDE_DIRS
  and
    LIBCONFIG++_INCLUDE_DIRS
  respectively, and can be used like this:
    include_directories(${LIBCONFIG++_INCLUDE_DIRS})

- Link to libconfig using variables
    LIBCONFIG_LIBRARIES
  and
    LIBCONFIG++_LIBRARIES
  respectively, by writing:
    target_link_libraries(${PROJECT_NAME}
      ${LIBCONFIG++_LIBRARIES}
    )

Also, add x64 solution platform and DebugStatic/ReleaseStatic solution &
project configurations, generating static libraries.
Reset output and intermediate directories and filenames to VS defaults.
Adjust preprocessor definitions (LIBCONFIG_STATIC / LIBCONFIG_EXPORTS,
_USRDLL) for the static configurations.

Re-generate scanner.[ch], add aux-build/test-driver, and update documentation.

Re-generate scanner.[ch], add aux-build/test-driver, and update documentation.

A value is recognized as CONFIG_TYPE_INT if it's only numbers, and
CONFIG_TYPE_INT64 if it's numbers followed by L or LL. However from a
user point of view adding L or LL is confusing. Instead, if the
integer is too big to fit in an int, return a CONFIG_TYPE_INT64. This
also avoid silent truncation.