Patched gio-2.0 VAPI files are no longer needed.

Signed-off-by: Maciek Borzecki <maciek.borzecki@gmail.com>

Implement mousepad plugin handler. The input in realized through AT-SPI.

Use Socket.create_socket_source() as it works for both TCP and UDP
sockets, but also does not use GDatagramBased interface, thus should
allow building on systems with older GLib versions.

A message ends with \n (newline), make sure that's the only extra
character we consume.

Replace named constructor Device.from_cache() with static method
Device.new_from_cache() to catch errors when loading from cache and
return null in such case.

IO receive from the mobile can be done synchronously without problems.

Calling Socket.create_source() will generate a C call to
g_socket_datagram_based_create_source(), and that returns null for TCP
sockets. Use a helper Socket.create_socket_source() method to workaround
this problem..

GLib.Socket implements GLib.DatagramBased, hence calling
Socket.create_source() results in a C call to
g_socket_datagram_based_create_source() insted of
g_socket_create_source() as one my expect. The former works only for
datagram based sockets (as implemented in gio/gsocket.c), hence
Socket.create_source() for TCP socket will always return null. To
workaround this, export g_socket_create_source() as
Socket.create_socket_source() method.

The change is expected to be a part of vala 0.34 release.

Due to gio-2.0 VAPI changes Socket.create_source() does not return a
SocketSource anymore, but a standard GLib.Source instead. Make sure that
we cast to SocketSource.

Prevent valac from producing colored output as it breaks Emacs compile
buffer handling and find-file support.

In preparation for DBus move the files to their respective directories.

VAPI is autogenerated based on GObject introspection. The problem is
that for gir to work a library must be run on the host, what obviously
brakes if cross compilation is used. As a workaround use the hand
generated VAPI and completely dump autogeneration.

Add hand crafted VAPI for mconn-crypt based on autogenerated version.

Part of namespac cleanup work.

This enables debugging with gdb. Without -g flag, the generated C code
does not have file/line set, hence gdb has no way of telling the Vala
code line the C code corresponds to.

Remove C files generated by valac in 'clean' target.

Use TCP keepalive for detecting of connection goes offline in a more
reliable manner.

Keep track of known device and dump their info to cache. This should
speed up reconnect after restarting mconnect. The rationale is that the
devices are likely to be connected to to the local network and use
DHCP and as such will most probably get the same address the next time
they appear. Hence, it should be ok to attemp connection the last know
IP address. If that fails, the discovery process will still take place.

Builds on COPR for F21 & rawhide