[cpp-threads] Thread API interface tweaks

[Peter Dimov]

Alexander Terekhov wrote:

> Ah. Okay. Now if you can make your future's operator() translate
> remote thread_cancel exception into future_canceled exception (when
> remote thread cancel wins the race),

future<> never translates, it throws whatever is passed to set_exception, 
but task<>::operator() can indeed be made to translate thread_cancel to 
remote_thread_cancel/future_canceled/whatever, removing the ambiguity.

> enable me to inspect the union
> without throwing/catching any exceptions, then I'll be totally happy.

I'm not sure what do you mean by that... can you elaborate?