[refactor] join(...tasks) & cancel(...tasks) should return join/cancel descriptors

[Andarist]

At the moment they are internally wrapped in all but actually the effect type should get preserved so internal refactor is needed.

At the same time we might consider changing those signatures to:

join([...tasks])
cancel([...tasks])

but I'm still on the fence with that, not sure which one is better - variadic arguments or a single array argument? any fresh thoughts on this one @restrry @shinima ?

[feichao93]

I'm inclining to single array argument.

The problem with variadic arguments I found is that join(singleTask) matches both join(...tasks) and join(task), and it is unclear that yield join(singleTask) should return taskResult or [ taskResult ].

It is clearer to use array argument: If you pass a single task object, then the result is a single result object; if you pass an array of tasks, then the result is an array too. Like in all or race effects, the result type (array or object) is determined by the argument type.

By the way, this issue had got some discussion in #1248.

[Andarist]

You have convinced me to array argument with simplicity of input arg type to output arg type mapping 👍

[mshustov]

@shinima would you like to work on it?

[feichao93]

OK. I'll get it done in a few days.

[feichao93]

Done in #1527