pluto: retry on empty private-dns-name from EC2

[etungsten]

Issue number:

Resolves #3363

Description of changes:

    pluto: retry on empty private-dns-name from EC2
    
    Use fibonacci backoff on requests to EC2 for fetching the private DNS
    name of the instance. Retry on both API failures and on when the private
    DNS name is empty.

Testing done:
Instance comes up fine. Directly calling pluto works:
On an instance where hostname is resource-based in the subnet:

bash-5.1# pluto private-dns-name
"i-088c99da0374c92a5.us-west-2.compute.internal"

On an instance where hostname is IP based:

bash-5.1# pluto private-dns-name
"ip-192-168-21-104.us-west-2.compute.internal"

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[stmcginnis]

Looks good!

[etungsten]

Re-evaluating approach since there is potentially a valid combination of VPC configuration and kubelet/aws-iam-authenticator configurations where a node can still join a cluster where private DNS name is empty.

[bcressey]

Re-evaluating approach since there is potentially a valid combination of VPC configuration and kubelet/aws-iam-authenticator configurations where a node can still join a cluster where private DNS name is empty.

@etungsten did you determine that no such valid combination exists or would ever have worked on Bottlerocket previously? That's my current understanding.

[etungsten]

There is a very fringe case where a user can potentially set up a bootstrap container to set their own hostname-override setting after pluto has set the hostname-override setting to "". After merging this, to maintain that same workflow, a user would just have to set "" or (a random string) for the hostname-override setting in userdata and let their bootstrap container do the overwrite,