Work around iOS7 memory trouble

[danzel]

Don't apply the contextId performance optimization when context==this. The optimization does nothing in this case anyway AFAIK.
Most of our event adders appear to have context==this according to testing.

Fixes #2122

[mourner]

Hmm, could you explain why you think it does nothing when context === this?

[danzel]

So the code works something like this:

addListener:
if we have a contextId, set up a map like (using {} for maps, [] for arrays):

events{event_idx}{contextId}[ ... ]
events{event_len}++

if we don't, set it up like:

events{event}[ ... ]

When an event is fired, if a context was provided when the event was added, fire the event in that context. If none was provided, fire it in the this context:
listeners[i].action.call(listeners[i].context || this, event);

So the no context list is essentially the context === this list.

If you run the L.MarkerClusterGroup 388 realworld example and log all places no context is passed in you get:
Path.SVG.js:165,173 - Which is taking advantage of context || this, so essentially it is context === this anyway.

Everywhere else passes a context.

Looking at it, we really shouldn't need the no context list, we should be able to set context = this in the no context place, but this breaks on iOS7 still (I hacked it up locally).

I have found a few other redundancies in the events code, will fix them up and add them on too.

[danzel]

Alright. I read the events code lots and did some tidy up and made this patch work correctly when removing event listeners.

[mourner]

Really great work Dave, thanks a lot for all the clarifications and improvements! I still can't wrap my head around the fact that the original code caused a memory leak in iOS7. Just don't understand why that happens. Any guesses?

[danzel]

I have absolutely no idea. Maybe maps/objects with a small amount of children have bad memory usage?
Still seems a bit extreme though!