if SNESSolve_VINEWTONRSLS was broken out early then make sure the DM is cleaned up and VI related information removed

Added SNESVIMonitorResidual() and -snes_vi_monitor_residual
Changed SNESMonitorVI() to SNESVIMonitor()

Moved DMDestroyVI() to before the SNESMonitor() calls so that any SNESMonitor callbacks have th

Added SNESVIMonitorResidual() and -snes_vi_monitor_residual
Changed SNESMonitorVI() to SNESVIMonitor()

Moved DMDestroyVI() to before the SNESMonitor() calls so that any SNESMonitor callbacks have the full DM and not the modified VI version

Reported-by: Ed Bueler <elbueler@alaska.edu>

show more ...

Make PCMG if used in VINEWTONRSLS use Galerkin coarse grids since rediscretization doesn't work for reduced space/active set methods
Fixed scaling of ex9.c example so that it works well with geometri

Make PCMG if used in VINEWTONRSLS use Galerkin coarse grids since rediscretization doesn't work for reduced space/active set methods
Fixed scaling of ex9.c example so that it works well with geometric multigrid

show more ...

Merge branch 'pr245/mapdes/dmcreateinjection-mat'

DMCreateInjection: use MatRestrict where coarse Vec is zero

Rather than pulling the scatter context out of the injection matrix and
using INSERT_VALUES, if we know the coarse Vec is already zero we

DMCreateInjection: use MatRestrict where coarse Vec is zero

Rather than pulling the scatter context out of the injection matrix and
using INSERT_VALUES, if we know the coarse Vec is already zero we can
just use MatRestrict directly.

show more ...

DMCreateInjection: return Mat rather than VecScatter

For (amongst others) FV discretisations, the natural injection may well
be cell averaging, which is not expressible as a VecScatter.

manual merge of conflict

removed excessive object reference that now leads to leaked memory

convert some PetscMalloc() to PetscMalloc1()
remove unneeded () in first argument to PetscMalloc/Calloc1()
fixes for MATLAB examples