Now all filters are stashed by default. In order to override this, the
parameter all=False needs to be passed, and the individual filters need
to be specified.

Individual filters can be stashed and popped by passing keyword
arguments to the respective functions. Additionally the filter state can
be obtained and set directly, with the get_filter_state() and
set_filter_state() functions.

This was not working due to filtering semantics modifications.

The edge index housekeeping was not performed on edges which were
deleted due to vertex removal.

This provides a more convenient way of deleting vertices or edges
according to a simple criterion.

Add support for v.{out|in}_neighbours(), for direct iteration over the
neighbours, as well as v.all_edges() and v.all_neighbours(). This also
properly exports the Vertex and Edge classes, with docstrings.

e.g., g.save("~/foo.xml")

i.e. g.degree_property_map("in") will create and return a vertex property map
which corresponds to the in-degrees of the vertices. This is useful for
temporarily modifying or getting an array of degrees.

A property map object has now a get_array() member which returns an internally
owned array pointing to the property values.

Reduce duplication of code for loading C++ modules.

This fixes a rather central bug, which causes duplicated indexes if
edges are removed and then new ones are added. Edge indexes are now
recycled as they are removed and then new ones are added. This still
guarantees O(1) complexity when adding or removing edges.

Avoids having to do

    g = Graph()
    g.load("foo")

each time. Now one can do simply,

    g = load_graph("foo")

Fantastic.

Use weakrefs to keep track of known property maps in graphs, and owner
graphs in property maps to avoid circular references, which may lead to
memory leaks.

This also fixes a vertex removal bug.

Graphs are now exactly copied, including filter state and filter
property maps.

Also, get rid of g.underlying_graph(), and rely on "private" variable
g._Graph__graph instead.

This commit removes the internal property maps from the GraphInterface
class, and makes all property maps external by default. The internal
property maps were moved to the python layer.

The whole histogram code has been redone, and the code has been
simplified. The three-point vertex-edge-vertex correlation has been
scrapped, since it's not frequently used, and would make compilation
even more expensive.

This also adds some missing files to the generation routine.

Now graphml files properly contain all the supported value types, which
are all perfectly preserved when read (floating point data is now saved
in hexadecimal format). Several other improvements were made, such as
the ability to read and write to python file-like objects.

It is also now possible to have arbitrary python object properties, and
store them persistently (which is done internally with the pickling
interface).

vector<bool> was totally abolished, since its implementation is quite
broken. See: http://www.gotw.ca/publications/N1211.pdf and
http://www.gotw.ca/publications/N1185.pdf Now a uint8_t (aka. char) is
used in graph properties instead of a bool.

Graph types can now be fully pickled (this may not be feasible
memory-wise if the graph is too large, since the whole XML
representation is dumped to a string before it is saved on disc).

This commit splits libraph_tool into different libraries:
 
   - libgraph_tool_core
   - libgraph_tool_clustering (*)
   - libgraph_tool_community (*)
   - libgraph_tool_correlations (*)
   - libgraph_tool_distance (*)
   - libgraph_tool_generation (*)
   - libgraph_tool_layout (*)
   - libgraph_tool_misc (*)
   - libgraph_tool_stats (*)

It also adds the python sub-module 'test', which provides extensive unit
testing of the core functionality. The core library is fully functional
and all test pass successfully.

(*) -> module needs to be ported to new refactoring, and does not yet build