[kube_state] add kubernetes_state.nodes.by_condition count

[xvello]

What does this PR do?

Add a count metric kubernetes_state.nodes.by_condition extracted from the kube_node_status_condition ksm metric.

Current known tag values:

• condition:
  • MemoryPressure
  • DiskPressure
  • OutOfDisk
  • Ready
• status:
  • true
  • false
  • unknown

Aggregation is done client-side, and we report zeroes for all statuses to avoid non-data graphing issues. That makes for 12 new timeseries per cluster. Node-specific state is visible via the existing service checks.

Motivation

Although the kubernetes_state.node.* service check allow to set alerts, we cannot graph them. This count could allow a user to graph their ready node count alongside their autoscaling group size for example. This count would not superseede the existing SCs, that offer node granularity.

Having node granularity on these counts would bring the no data issues we have with kubernetes_state.pod.ready when churning.

Versioning

• Bumped the check version in manifest.json
• Bumped the check version in datadog_checks/{integration}/__init__.py
• Updated CHANGELOG.md. Please use Unreleased as the date in the title
  for the new section.
• If PR impacts documentation, docs team has been notified or an issue has been opened on the documentation repo

Additional Notes

Anything else we should know when reviewing?

[mfpierre]

LGTM