No more need to run soffice with LIBO_TUBES set.

Pass const rtl::OUString &s around instead of random pointers.
Allow sc/ to build and work without telepathy enabled.
Add aforementioned demo mode hack.

This is no longer needed since the switch to using GDBus for the tube
connection.

I added these to the recently-released telepathy-glib 0.18.0 mainly to
let us simplify the code here. With this new API, tp-glib takes care of
waiting for the state changes, creating the GDBusConnection, etc.

It already was, actually. :)

GDBus is the future. And TpDBusTubeChannel will ultimately give us a
GDBusConnection, so...

This makes sense, because the flow for Accept() and Offer() is basically
symmetrical, and then conference becomes solely responsible for the
handling of channels once they've been requested/given to us.

While we're at it we also fix both code paths to work correctly if the
state becomes Open before Accept/Offer returns (which is possible).

initialize with same mode as collaboration

relict from testing era, not needed anymore

* "passing argument 1 of ‘gtk_message_dialog_format_secondary_markup’ from
  incompatible pointer type"
  * dialog was in fact constructed as GtkMessageDialog using
    gtk_message_dialog_new_with_markup(), so use it as such
* "passing argument 1 of ‘gtk_widget_destroy’ from incompatible pointer type"
  * use GTK_WIDGET() on dialog GtkMessageDialog*
* "passing argument 1 of ‘g_object_get_data’ from incompatible pointer type"
  * use G_OBJECT on dialog GtkWidget*
* "passing argument 1 of ‘gtk_dialog_set_response_sensitive’ from incompatible
  pointer type"
  * use GTK_DIALOG on dialog GtkWidget*
* "passing argument 1 of ‘gtk_message_dialog_set_image’ from incompatible
  pointer type" and "passing argument 1 of ‘gtk_dialog_add_buttons’ from
  incompatible pointer type"
  * use GTK_MESSAGE_DIALOG() on dialog GtkWidget*
* "passing argument 1 of ‘g_object_set_data_full’ from incompatible pointer
  type"
  * use G_OBJECT() on dialog GtkWidget*:x
* "passing argument 3 of ‘g_signal_connect_data’ from incompatible pointer
  type"
  * use G_CALLBACK() on dialog_response_cb
* "passing argument 1 of ‘gtk_window_set_skip_taskbar_hint’ from incompatible
  pointer type"
  * use GTK_WINDOW() on dialog GtkWidget*

How awful C is :-/  GtkMessageDialog is a GtkDialog is a GtkWindow is
a GtkWidget is a GObject, but still ...

This is the bug which breaks the first request after the offerer signed
in.

To record a screencast, I wanted to get two instances of LibreOffice
running in the same session. But the Client object(s) claim D-Bus names,
so initializing the second instance failed.

This patch changes things so that if LIBO_TUBES=master, no Client
objects are created. As a result, only the slave can receive files.

(And also signalify receiving files.)

Yes, this is in addition to the existing TeleManagerImpl singleton. This
class needs to be properly split in half: one Manager part from which
the UI can request new sessions and which signals the appearance of new
incoming sessions, and another Session part representing the shared
editing session (which in turn owns one or more Conferences, which owns
exactly one tube, as now).  The Manager will dispatch incoming files to
the appropriate Conference by UUID or similar.

But for now, Michael is opening a new window with the received file,
so we want incoming and outgoing events to go to both windows so that it
works well enough for a demo.

This doesn't actually seem to have any effect on the test passing or
failing.

I think I commented these out while hunting stack corruption.