Some backends create nodes in their lower_for_target phases already.
These nodes are placeholders for real backend nodes later, in contrast
they do not have register requirements or backend attributes. Simplified
this by allowing custom dumpers for them and leaving out the backend
node structs.

This fixes opt/fehler202.c.

This is a merge, because it should be considered as a single commit
and libFirm might be broken between the various commits of the branch.
However, there are so many changes, that i do not want to squash
them all into one mega-commit.

API change: removes set_irg_outs_inconsistent and get_irg_outs_state
Use set_irg_state and is_irg_state instead!

Converted: ifconv, boolopt, combo, fpvrp, ldst, code placement,
jumpthreading, deconv, parallelize mem, loop, scalar replacement,
cfopt.

This is a basic implementation of http://libfirm.org/PhaseManagement

Introduces remove_unreachable_blocks as an extra API function,
which does a subset of optimize_graph_df.

Cheering up Tron ;)

returns_twice handling was broken in the ia32-backend (the only real
beabi user) for a long time now. Just noone noticed because our
frontends lacked the special treatment to set the flag correctly.
Turned out setjmp/longjmp on ia32 save/restore the registers anyway
so not doing anything is fine here.

They might be the target of an ijmp afterall.

This fixes backend/float_ret_sse.c with -O1.

Implement a generic way to lower compound-call-arguments.
Parameters are prepared on the stackframe and passed as a pointer
to the function.

This should fix fehler200.c

This should also fix compound returns in combination with
-msoft-float on sparc

Softfloat library does not support long double anyway.

(Hannes ancient technique based on the Proj-mode is deprecated)