This is done by creating a fake 'virtual' environment at a fake
point and running the symbolizer render code. The actual render
is saved in a thunk for after the group layout has been done.

This renders the saved information from previous calls to the
bounding box extraction code, offset by some amount which should
be determined from running the `placement_finder`. Note that this
doesn't implement that bit, just the rendering.

Create a group_symbolizer_helper for group placments, and extract some code from
text_symbolizer_helper into a base class to share with group_symbolizer_helper.
Also, move tolerance_iterator into its own header file. Use helper in
process_group_symbolizer to find placement positions.

Segfault was due to `glyph_position` structs keeping a pointer to
`glyph_info` objects which went out of scope at a different time.
The (rather ugly) fix for the moment is to copy that information
into the thunk object.

Move a lot of processing into a common process_group_symbolizer function.
Also, extract column collection out of process_group_symbolizer function.
This will reduce duplication needed for other renderers.

The first test shows how an obstacle under the group being displayed
will cause the whole group to not display. Above is an replica of
the same layout using non-grouped symbolizers, which shows that
only the parts which intersect the obstacle will not be drawn.

The second is a line placement test, which just tests placement
of groups on a line with variable spacing.

The third is a test of repeat key in group symbolizer.
The test shows two items with different repeat keys ("foo" and "bar")
and rows of other group symbolizers sensitive to either one. The
reference images show that the "bar"-sensitive rows are displaced
by the "bar" item, but not the "foo" item.

This uses the renderer_common header to do most of the heavy
lifting, but otherwise is very similar to the AGG renderer
implementation.

Add cairo ref images for group symbolizer tests.

Debug symbolizer is useful, and used in a bunch of tests. This
adds debug symbolizer support for Cairo to make it closer to the
capabilities of AGG.

Adding debug symbolizer for Cairo meant that red boxes appeared in
many of the visual test outputs. This commit replaces them with
the output, after visual inspection. They should now be closer to
the output of the AGG test cases.

Note that the "implementation" for SVG renderer is the same as
for other symbolizers there - i.e: empty.

This uses common process_group_symbolizer for most of the work.
Add reference grids for visual tests.

Boost::Python interface, and added a test to check that it works.
Looks like it fails at the moment due to other changes - a lot of
the python symbolizer tests show the same thing.

…-symbolizer

Conflicts:
	include/mapnik/attribute_collector.hpp

…-symbolizer

Conflicts:
	src/text/placement_finder.cpp
	tests/visual_tests/test.py