Cleaner method to get IS in solver

[gthyagi]

Replaced the use of the LocalSection method (complex for loop in cython) for obtaining an IS with the CreateSubDM approach (with petsc4py).

[gthyagi]

@lmoresi and @julesghub, could you please point me to examples where I can test the current implementation for obtaining the correct IS? Thank you!

[julesghub]

As discussed a simple 4x2 mesh with velocity, pressure and temperature fields.
Where you could count the indices by hand.
Steady state solve for T.
Stokes solve for v,p. You could even make the RHS of stokes weakly dependent on T.