The executables are no longer removed when the unittest fails.

Newer Linux systems compile ar with --enable-deterministic-archives,
which conflicts with the u modifier.

Also improve the comment a bit

-T means something entierly different for non-GNU ar.

Previously we had an AR commandline that replaced old object files with newer
ones in the archive. This however fails when we remove files from the
project, because the previous objects in the archive were not removed
(and often caused missing symbol errors when linking later).
So now I remove the archive each time before creating it, to make up for
the speed loss use the "thin archive" option. This is not a good idea
for a release/installed version of libfirm, but for that we still use
the automake build anyway.

This mainly allows us to use external templates like the ones used for
homepage.

There was a lot of duplicated code, some changes in the templates allows
to share the two generators.

Before this change the .so file would not mention libm as a dependency,
so users had to explicitely link -lm when linking -lfirm.

This seems to be done by a lot of projects out there while our previous
Q= seems to be exotic, so we additionally support V=1 now too.
(based on a patch from lu_zero@gentoo.org

start with some unittests for raw bitsets.

This is a major Makefile refactoring:
differentiate between:
- source files in $(srcdir)
- generated files in $(top_builddir)/gen
- object files in $(top_builddir)/$(variant)/ aka $(builddir)
- documentation in $(top_builddir)/doc

the templates are now friendlier for usage outside of libfirm

Templates are put in separate files now, making the main script more
generic in nature

It has been empty for nearly all systems. People who used to put stuff
in config.h can still create a config.h on their own and inject
-include (gcc) or /FI (msvc) into their CPPFLAGS.

Since we have two fp-vrp.o files, the -u argument of ar sometimes (on
rebuilds) missed one. As -u is not necessary anyways, we now include both.