Cherry-pick of PR #14738 to 7.x branch. Original message:

This PR tries to move around code in a way such that each resource has its own interface implementation and the autodiscover relies on a resource parameter to determine what the watcher to spin up and how to create hints. We also now add scope as a parameter to define if Beats is running either at a node level or at a cluster level.

metricbeat.autodiscover:
  providers:
    - type: kubernetes
      kube_config: ${HOME}/.kube/config
      resource: service
      hints.enabled: true

output.console.pretty: true

The above configuration would enable discovering service objects across the cluster.

metricbeat.autodiscover:
  providers:
    - type: kubernetes
      kube_config: ${HOME}/.kube/config
      resource: pod
      scope: cluster
      hints.enabled: true

The above configuration would allow doing pod discovery across the cluster.

closes #14044 and #10578