[dune-pdelab] Vector and Matrix backends for PETSc

[Steffen Müthing]

If you want to look at the changes I did to the interface to shoehorn PETSc in there somehow,
the three relevant commits are:

r1607: Changes the way the local assembler accesses the matrix (the user perspective)
r1613: Makes the ISTLMatrixBackend compatible with that new access pattern
r1619: Not really important, but makes sure the default GridOperator actually does the right thing

Steffen