docs: add note about chord and result backend

[hrbonz]

Added a note at the beginning of the chord to point at the necessity to enable result backend for tasks inside a chord.

Proof of stupidity/lack of clarity in the docs/RTFM: #1857

[ask]

Thanks! It could also be an idea to raise an exception if any of the tasks in a chord has ignore_result=True.

For 3.2 I want result to be stored on a task invocation basis so that you can call task.apply_async(reply='redis://localhost') so that the caller can specify if the state is required or not, and that way chord can ignore the setting and just make it work anyway.

[hrbonz]

I was thinking about it this afternoon, it would be great to be able to pass ignore_result to apply_async() or in subtask(). I'm sure there is a good reason for that behaviour, I'm still not comfortable with the source.

On another note, I found the chord / head / body a bit unintuitive, I guess head / callback is easier to understand. There are also places where callback is used and others where body is used. I wanted to rewrite some of the chord bit to make it that way.

[ask]

The terminology is from Cω (http://research.microsoft.com/en-us/um/cambridge/projects/comega/doc/comega_langspec_concurrencyconstructs.htm)

[hrbonz]

Doesn't mean it's the best or easiest way to name those 😄
I'll propose a PR to make the canvas page a bit easier for newcomers, it took me a bit of time to get it properly.

[floer32]

👍 this confused me as well. (Admitted beginner.)