CogLauncher should probably not be a GApplication

[igalia]

CogLauncher should probably not be a GApplication

[aperezdc]

This idea has been in my head as well, because it forces uses of libcogcore to have a GApplication even if a particular application does not want one or cannot use one for some reason. What I would do is:

• Split the functionality that is independent from GApplication out of CogLauncher into a new class, let's say CogView (if somebody can think of a better name, I'm all ears).
• Move plug-in handling into the new CogView class. This way code that uses libcogcore does not need to duplicate code that now lives in cog.c.
• Keep the CogLauncher class around, but make it use the new CogView class.

I like the idea of keeping a basic GApplication subclass around, because for cases where using it can be used it will avoid third party developers having to write a good chunk of non-trivial code. One of my mental goals in this regard is that writing a very basic launcher using libcogcore should be possible in under <500 LOC (currently cog.c is 368 LOC).