PATCH: last DFP patch

[Ben Elliston]

> This patch completes the libgcc changes for decimal floating point
> support.  It allows users to include the DFP arithmetic routines in
> libgcc on selected targets (x86, x86-64 and PowerPC GNU/Linux) when
> GCC is configured with --enable-decimal-float.

> 	* config.gcc: Include t-dfprules in $tmake_file for i?86-*-linux*,
> 	powerpc-*-linux* and powerpc64-*-linux* configurations.
> 	* mklibgcc.in: Add logic to handle $DFP_ENABLE and $DFP_CFLAGS.
> 	Compile dfp-bit.c from the source directory using -DWIDTH and
> 	$DFP_CFLAGS to control the compilation of that file, rather than
> 	generating width-specific versions in the build directory.
> 	* Makefile.in (libgcc.mk): Pass down DFP_ENABLE and DFP_CFLAGS.
> 	(LIBGCC_DEPS): Replace $(D32PBIT) et al with config/dfp-bit.[hc].
> 	* config/t-dfprules: New file.
> 	* config/i386/t-linux64 (DFP_ENABLE, DFP_CFLAGS): Define.

DJ has approved the *.in changes.  David Edelsohn has approved the
PowerPC-specific changes.  I've committed the pieces approved thus
far.  Thanks,

Ben