Reference counts of newly created arrays must always be incremented, and
not borrowed().

This prints useful library information.

We will rely on sphinx + docstrings instead.

Import numpy and scipy before everything to avoid weird segmentation
faults depending on the order things are imported.

Python types cannot be checked in parallel. This commit forces the
number of threads to be 1 in such case.

This changes degree_selector(). Now all properties should be tested, not
only scalar properties. Later it can be decided which properties are
valid.

random_graph() now uses a modified algorithm for generation of
correlated graphs, which is more efficient. Instead of giving a function
which returns a sample of the correlated target degree, the user must
give a function which will just compute its probability. This
probability will then be used to choose the edges.

This also simplifies the calculation of the coefficient.

This modified several docstrings and introduces a quick-start guide.

The generators from boost::random seem to have a bug which causes them
to be biased. The generators from tr1::random seem to be in better
shape.

See: http://thread.gmane.org/gmane.comp.lib.boost.user/48006

Now the parameter "all" is ignored, if any of the other parameters is
True.

The following is now possible:

    g = Graph(directed=False)

Instead of the only way previously:

   g = Graph()
   g.set_directed(False)

The following conversions are now possible: int(v) and tuple(e), where v
and e are Vertex and Edge instances, respectively.

Now paths are weighted according to their trust value until the last
hop, and paths are not counted twice.

Vertex/edge colors are properly plotted even when value range is zero
lenght. The behaviour of the 'fork' option was slightly modified.

graph_draw() won't work if gv is not available, and just a warning is
printed... This makes graph_tool.all work when gv is not available.

Fix bad_any_cast bug when passing a weight property.

This fully documents the centrality module.