Test Runner and KubernetesEngine New Design

[tylergu]

Proposal

To design a runner interface to support different implementations of runner.

A test runner takes a CR as input, applies it to Kubernetes cluster, waits for the system state to converge, collects and returns a Snapshot of the system state.

Acto currently implements a local runner, which uses a local Kubernetes cluster. If we have a uniform runner interface, we can have a distributed runner implementation which can utilize multiple machines at same time for single test runs.

Ideally this is to be distributed via a framework, e.g., Ray. Implementing a distributed system would be a rabbit hole of engineering work.

Migrated from closed PR: #235 #265