Merge lp://qastaging/~apinheiro/unity/atk-selection into lp://qastaging/unity

This branch solves the problems with all the related bugs. All three were too tied to create three different branches, so I concluded that it was better to just work on a branch.

Anyway, I have a problem. I was not able to test it. On a normal run, it crashes when the launcher icon accessible object tries to connect to the new signal that I added to the Launcher. This is weird, as it works on the custom unit tests that I created (on unitya11y). This is the reason I added Mirco here, to check if the signal is wrong.

Anyway, I fear that it is a known bug. Using gdb on this crash I get this backtrace:
rogram received signal SIGSEGV, Segmentation fault.
0xb76070a9 in std::_List_node_base::_M_hook(std::_List_node_base*) () from /usr/lib/libstdc++.so.6
(gdb) bt
#0 0xb76070a9 in std::_List_node_base::_M_hook(std::_List_node_base*) () from /usr/lib/libstdc++.so.6
#1 0xb76d1c4d in sigc::internal::signal_impl::insert(std::_List_iterator<sigc::slot_base>, sigc::slot_base const&) () from /usr/lib/libsigc-2.0.so.0
#2 0xb76d1ce7 in sigc::internal::signal_impl::connect(sigc::slot_base const&) () from /usr/lib/libsigc-2.0.so.0
#3 0xb76d1f78 in sigc::signal_base::connect(sigc::slot_base const&) () from /usr/lib/libsigc-2.0.so.0
#4 0xb509fd4d in connect (accessible=0x8de9920, data=0x91efd0c) at /usr/include/sigc++-2.0/sigc++/signal.h:2659
#5 unity_launcher_icon_accessible_initialize (accessible=0x8de9920, data=0x91efd0c

And this _M_hook thing is a bug already reported:

https://bugs.launchpad.net/ubuntu/+source/nux/+bug/717600

So, if I'm proposing this merge, is more a gentle request to the people: could you review my code to discard any stupid mistake in my code? Could you test (Rodrigo?) this branch to discard something specific on my environment? Thanks