Change upgrade logic for worker nodes

[mrajashree]

Worker nodes should be upgraded in user configurable batches. Users can specify maximum number of nodes that can be unavailable during the upgrade. The upgrade will follow a sliding window approach based on maxUnavailable value users specify.
The nodes will be cordoned before upgrade. Users can also optionally drain the node.
Nodes will be uncordoned after upgrade.
Users can specify maxUnavailable and drain input through cluster.yml

upgrade_strategy:
  max_unavailable: 1
  drain: true
  node_drain_input:
    ignore_daemonsets: true
    force:
    delete_local_data:
    grace_period:
    timeout:

[soumyalj]

Tested with rke cli - v1.1.0-rc5
Some of the tests executed :

1. Verified that worker nodes are upgraded in batches . maxUnavailable can be configured which defines the number of nodes that get upgraded at a time.
2. Verified worker node upgrade by specifying the drain=true

 upgrade_strategy:
  max_unavailable: 3
  drain: true
  node_drain_input:
    force: False
    delete_local_data: false
    grace_period: 30
    timeout:120

3. Update cluster and specify drain=true option and change kubelet flag or change any parameter other than K8s version.
4. Drain failure during cluster update Drain failure during cluster update
5. Add a worker node in rke file and then upgrade specifying maxUnavailable field
6. Remove a worker node in rke file and then upgrade specifying maxUnavailable field

Worker node upgrade succeeds in all the cases as expected. The nodes have updated images for worker node components after upgrade (when there is a change in image)