Now we can call functions that need to access GPU resources explicitly
on the draw thread.

This lets us get rid of m_autorelease and m_destroy and we’ll be able
to implement init_game/init_draw later, for stuff that must happen
on the render thread.

We need to ensure the inter-thread messaging queues are empty (more
specifically, not locked) before destroying them. The best way to ensure
that is to explicitly wait for thread termination.

The scene now owns its renderer, so there is no need to keep a list
of existing renderers and to call Renderer::Get().

This was not a very smart class. We replace it with a bidirectional map.

We used to have lol::map::empty() to empty maps, but in std::map
this method is called clear(). It sounds safe to harmonise between
our types and have lol::array::clear() too.

From now on, msg::debug will not display anything, unless this is a debug
build, or the LOL_DEBUG environment variable is set.

This allows actors to perform draw commands in TickDraw(), even
if it’s not exactly what we want in the future.

Not all programs support this feature for now, but that’s because they
don’t support the scene framework either, and will need patching. Also
the default postprocess is deliberately exaggerated for now.

Now TickDraw() can be called several times per frame, because we
may have several scenes. This is expected, but it will require
some adjustments later.