Cancellable

[raphael-proust]

Fixes #782

This PR provides a new cancelable : 'a t -> 'a t function such that cancelable p is a promise p' that is cancelable even if p isn't.

I introduced a new constructor in how_to_cancel. I'm not 100% sure it is necessary but I couldn't find a way to do without it.

[hcarty]

The concept makes sense, but it would be nice to give this a clearer name since it doesn't make a promise cancelable. Rather it wraps a promise in a new, cancelable promise. Maybe wrap_in_cancelable?

[raphael-proust]

@hcarty I agree that the name is not really good as is. The type of the function hints that it wraps the promise (rather than modify the promise in place) but the name should also hint at it.

I like wrap_in_cancelable. I tend to like past participles for this kind of things – e.g., how Python has sort (for essentially 'a list -> unit) and sorted ('a list -> 'a list). And that would mirror protected nicely too. But I can't find a good past participle.

Anyway, I'll rename it to something like wrap_in_cancelable if that's the best suggestion I get.

[raphael-proust]

So I actually made it point to the right test (thanks for spotting the error) and there's a bunch of failures where tests raise CamlinternalLazy.Undefined. I'll investigate the feature and maybe I'll have to go without Lazy and reintroduce the Cancel_this_promise_and_propagate_to_one.

[aantron]

I think, in the worst case, you should be able to define the function using higher-level helpers like on_cancel. I could be wrong. But if that's correct, I suggest to go that route rather than introducing a new internal constructor.

[raphael-proust]

I implemented it with on_cancel for backwards cancellation propagation and on_any for forward resolution propagation.

It doesn't fit with the protected/no_cancel syntactically because of the order the primitives are declared in, but it works.

[aantron]

Thanks!