You now only have to specify 1 type and the size of the access instead
of 2 types. Also cleaned up some code/comments in the process.

The matrices generated by execfreq tend to be not well conditioned.
Therefore, it is advisable to use a numerically stable algorithm, such as
QR decomposition.

Besides, QR decomposition can be used to directly get a matrix' null space,
which is what execfreq actually needs.

Only blocks which have a backward jump will be part of the linear system,
because the execution frequencies of other blocks can be computed by simple
substition from the predecessors.

This speeds up the execfreq algorithm and finds a solution more often, but may
still have problems with some ill-conditioned matrices.

- It was only available in the ia32 backend and broken there. As the
  feature is rarely used, I remove it for now instead of going through
  the trouble of fixing it.

This should fix the 186.crafty problem.

The previous approach put the irg link into the block attributes, so
while saving memory an extra indirection was necessary. Some testing
with valgrind massif+callgrind revealed that a direct link increses peak
memory usage by ~1% while also increasing runtime by ~1%. More
importantly the code gets a bit simpler and some strange corner cases can be
removed with the direct link.

Make the list scheduler a set of helper functions, instead of a complex
design with a set of callbacks.
Simplified list scheduler code.

In their current state they produce very bad code and are just a
maintenance burden.

This fixes backend/sparc_segfault.c.

This fixes a subtle bug where the delay slot filler would move a
spill/reload across a MemPerm although they both share a spillslot.

The global analysis has quadratic worst-case complexity in the number of
graphs.  At least if alias analysis is disabled anyway, this can be
avoided.