feat: Add extended TaskClient implementation

[matyascimbulka]

Implements #29.

This PR implements extended TaskClient for the Apify orchestrator. However the Typescript types aren't cooperating with me here. I have tried to make it as close as possible to the already existing extended ActorClient.

Do you have any idea how to make it work? Or what I'm doing wrong?

cc: @gullmar @JuanGalilea @kazadoiyul

---

EDIT:
I have moved the runName parameter into the options object to satisfy class inheritance. I have also moved some common code between actor-client.ts and task-client.ts to utils and consts.

Finally I'm not sure how to test it apart from writing unit tests. @gullmar any help with testing would be appreciate.

[JuanGalilea]

Due to restrictions of inheritance you cannot do it that way with .start and .call. Someone expecting a TaskClient should be able to recieve an ExtendedTaskClient and be none the wiser, but since it takes a different input for call and start it would have issues.

You can deal with this in 2 ways, depending on your requirements:

• Move runName to options: This allows you to remain as a subclass of TaskClient but has the drawback of runName becoming optional. You can return it to the user as part of the output (if they need it)
• Not inherit and just use TaskClient under-the-hood: This allows you to keep runName as required (and have whatever interface you want) but you lose the being a subclass of TaskClient and all the benefits that entails (mainly that it is a drop-in improvement).

PS: after some discussion with @matyascimbulka it seems like the first option is the best suited for this problem, along with some sane generation of runName (taskName + some random chars) in case it was not specified by the user.

[matyascimbulka]

I think this PR is ready for review.

Thank you @JuanGalilea for your help. I have decided to move runName into options and do a runtime check to enforce it. In the end I don't think that random (or semi-random) generation of the runName wouldn't work since you need to be able identify the run on subsequent runs (e.g. migrations).