Fix more warnings

See merge request !195

It's the level of a vertex in a grid, so supposedly it can never
have negative values.  However, it is used to index arrays,
and some compilers warn about indexing arrays with signed types.

Fixed warning "Empty_Rule is defined, but unused" in for 2D and 3D

See merge request !194

Remove a few warnings

See merge request !193

This fixes a few 'uninitialized value' warnings.

Link dunecommon to duneuggrid lib

Closes #55

See merge request !191

[dddtypes] Fix missing include for std::size_t

See merge request !190

Remove ENABLE_UG Flag from cmake files

See merge request !189

Changelog Added entry about All_All communication on facets.

See merge request !185

Bump version to 2.9-git

See merge request !184

Fix some warnings regarding unused results and printf format

See merge request !182

Update list of authors

See merge request !180

Implement the All_All communication interface for facets

See merge request !178

Drop XDR support

See merge request !177

Fix Debug mode

Closes #35

See merge request !160

If I understand it correctly (no documentation in sight), this mechanism
was able to detect whether pointers into the UG3 memory heap were stale,
i.e., did not point to a valid object anymore.  This was used in various
places to verify the data structure integrity. As we have removed the
custom memory heap a long time ago, the HEAPCHECK mechanism cannot be
used anymore (at least to my knowledge you cannot check for pointer
staleness in the system heap).  Nobody noticed this, because the
corresponding assertions were never enabled.  Enabling them leads
to occasional program termination with spurious error messages,
and therefore I herewith remove HEAPFAULT and HEAPCHECK.

HEAPFAULT is some form of consistency check, but I don't really
understand what it is supposed to do.  It may just be a left-over
from the days when UG3 implemented its own memory heap.

The HEAPFAULT in question here triggers an assertion failure
in test-parallel-ug.

It asserts begin < end for a pair of iterators, but the code
right below it handles the cases begin==end, so the assertion
must be wrong.  Nobody notices because apparently nobody ever
enabled the assertion.  Once it is enabled it fires in the
dune-grid standard tests.

The previous way to enable it didn't actually work.

There is an assertion in GetDomainPart that fails when it is actually
enabled.  While this makes it likely that there is a bug somewhere,
the 'part' information is never actually used Dune UGGrid.  I therefore
simply disable the assertion, expecting that this code will be removed
eventually anyway.