implementation of generic scale from zero

[AxiomSamarth]

What this PR does / why we need it:
This PR implements the generic scale from zero logic.

Which issue(s) this PR fixes:
Fixes #27

Special notes for your reviewer:

Release note:

Gardener Autoscaler now supports generic scale from zero. Traditionally, nodes are scaled by Autoscaler by generating node template for a new node that would accomodate the pending pods. This node template is generated either by referring to the node spec of the already existing node in the nodeGroup or by referring to the cloud provider specific catalogue which provides the node details like `cpu`, `gpu`, `memory`, `region`, `zone`, `instanceType` etc. The latter scenario is the scale from zero scenario where there are zero nodes in the `nodeGroup` that is expected to scale up and so far the existing logic limited this feature to AWS and Azure only.

Now, with the introduction of `nodeTemplate` property in the `MachineClass`, the scale from zero feature has been extended generically across all providers. 

[AxiomSamarth]

The changes are tested with the below steps on AWS, OpenStack

1. Create a Shoot with a worker pool spanning across multiple zones with minimum number of machines equal to 1 and maximum number of machines equal to number of zones the worker pool is spanning.
2. Edit the MachineClasses on the Shoot by adding the expectedNodeAttributes field.
3. Run the Cluster Autoscaler locally by trying to schedule pods that become unschedulable to witness the scale up happening in other zones which has machineDeployments with 0 replicas.

[prashanth26]

Changes look good to me.

• However, improve the release notes
• to provide more details on how to make use of it as end user
• As well about deprecation of current scale from zero logic.
• Finally adding some tests will help

/lgtm otherwise.
/needs second-opinion
@himanshu-kun

[gardener-robot]

@prashanth26 Label needs/@himanshu-kun does not exist.

[AxiomSamarth]

@prashanth26 Sure, Prashanth. I will incorporate those suggestions.

[gardener-robot]

@himanshu-kun, @hardikdr You have pull request review open invite, please check

[himanshu-kun]

Thanks for the changes, looks good , I will test myself once as well. Adding a unit test would be good.

[AxiomSamarth]

The machine catalogue is updated with the new data types. However, new files are created in the mcm package (as changing in the original location under aws and azure packages would lead to CA build failures). The changes in the mcm_manager.go has been done accordingly.

Tested for backward compatibility. Works fine.

The CRD & API changes from the new MCM release should be vendored in this PR. Post which the CA build will go through successfully.

[himanshu-kun]

Thanks for the neat PR Samarth
As suggested I think we can change variable name MemoryMb to Memory as it could be misleading . You might have to generate the catalogues again in that case.

Also as discussed by us of not fetching region and zone from expectedNodeAttributes now, I think those changes will get incorporated in your next push. Just commenting as a reminder

[himanshu-kun]

Minor change requested and discussion

[himanshu-kun]

Thanks for the neat changes, just some minor picks to improve error and logging messages

[himanshu-kun]

lgtm

[AxiomSamarth]

Merging the changes as decided.