Until now, using a list as prefix/suffix didn't work as expected,
because the corresponding method for retrieving the value was called
multiple times, and each time, the next icon would be retrieved.
Changed the logic inside the theme to only update the indices every time
that an actual update was happening.
According to the unit tests, at least, the old functionality is back
again - with the additional i3 block abstraction in output in place.
Also, pango support is temporarily removed again and will be
re-implemented based on the new architecture.
Add a way for themes to specify custom separators. Doing that, make
nicer interfaces for drawing "supplementary" components (separators)
for widgets and generalize the attribute retrieval within the theme.
Add two new parameters: theme and iconset
Add a placeholder class core.theme.Theme, an instance of which is passed
in to the i3 output object (which is the only object that should ever
have need of the theme, hopefully).
To make it easier to update individual modules, separate the call to
update() and the call to actually drawing the status.
Additionally, this avoids the "side effect" of updating when drawing the
status line.
Add a (half-finished) input library, that for now simply allows
registration and triggering of events.
As next steps, the trigger will happen as part of a separate thread that
reads input events.
Additionally, invoking commands via a execute() will be supported.
Thirdly, there is need of a way to selectively update the affected
modules (widgets), which should be possible given that the event
contains both the instance (widget ID) and name (module name).
Parse '-p' parameters and make them accessible via 'parameter()' from
within a module (basically, restore the previous configuration
facility).
Still TODO: Add support for configuration files.
The core.output module now manages the list of modules and retrieves the
widgets inside draw() itself. That way, details of drawing/updating
widgets are not visible from the outside anymore.
Add generic "draw()" method that redirects internally to the actual
calls. These can now produce JSON, which is nicer because:
1. Easier to use during testing
2. More flexible
3. Centralizes printing (somewhat)
Still, the "suffix" concept isn't really nice, but so far, I have no
better approach.
To maintain backwards compatibility (and because I think it's an OK
design choice), keep the widget concept (a single module can produce
multiple widgets).
