Autoscaling actor groups in runtime

[Relrin]

As the part of this PR, I would to introduce the changes that cover needs in autoscaling actor groups in runtime. The initiative was described in the #155 issue.

The idea

Before starting with the example, any developers needs to enable the scaling feature flag which using for getting an access to desired API, collecting statistical information from actors, etc.

Then, we can import the resizer, and declare the rules around how much actors acceptable (minimal, maximum) for the each actor group. It will be look something like this:

children
    .with_redundancy(3)                                // Start with 3 actors
    .with_resizer(
        Resizer::default()
            .with_lower_bound(0)                       // A minimal acceptable size of group
            .with_upper_bound(UpperBound::Limit(10))   // Max 10 actors in runtime
            .with_upscale_strategy(UpscaleStrategy::MailboxSizeThreshold(3)) // Scale up when a half of actors have more then 3 messages
            .with_upscale_rate(0.1)                    // Increase the size of group on 10%, if necessary to scale up
            .with_downscale_rate(0.2)                  // Decrease the size of group on 20%, if too many free actors
    )

The resizers will take care about reacting on collected stats (currently: active actors and the average mailbox size) and will ask a Child instance to scale up or down the actor group when it necessary in runtime.

Checklist

• added example
• tests are passing with cargo test.
• tests and/or benchmarks are included
• documentation is changed or added
• commit message is clear

[vertexclique]

This is state of the art. Just ornamental things. The rest is looking amazing. Awesome work. Thanks.