Fix regression on the nil delegate #706

[RuiAAPeres]

As discussed here #724

[RuiAAPeres]

The cell sizing is working (with a nil delegate) along side with the table:didSelectRowAtIndexPath: (with a non-nil delegate).

[appleguy]

Is this change actually necessary, given that it will be replaced / overwritten below anyway? If it is necessary, could you add a comment that explains why?

[appleguy]

@RuiAAPeres Thank you for looking at this! I would love to land this but am not yet comfortable as I don't understand the significance of one of the core changes here, moving the = nil outside the conditionals. Once I understand that, and if applicable (e.g. it is subtle reasoning and critical) it is commented, we'll get it in.

[appleguy]

@RuiAAPeres ping - I'm going to branch for the next release soon (looks like we'll call it 1.9 — representing confidence in improvements for all 1.x surface area, but some fluidity remaining for 2.x APIs). This is not /essential/ for the branch, but nice to have.

[RuiAAPeres]

@appleguy sorry for the delay. About the changes, they allowed me to be able to fix both the cell height's issue (when no delegate is assigned), enable table:didSelectRowAtIndexPath: and make the tests pass. I am unsure how/why setting _asyncDataSource = nil; before assigning it to a proper delegate makes it work, but the reality is that it does (I am suspecting there is a bug in UIKit). On my current project I am using my own branch, and things are working as expected.

I do understand, that you need a more convincing proof, but this week, I don't have time to create a project that:

1. Is able to reproduce this crash without any of my changes and then with my changes prevent the crash.
2. With a nil delegate, cells will have the proper size defined (according to calculateSizeThatFits).
3. With a non-nil delegate, table:didSelectRowAtIndexPath: will work.
4. The tests pass.