Return Node instances from GetNodes

[danieljanes]

Issue

Description

The Driver API returns individual int node IDs, not full Node instances. When scheduling a Task, one has to construct a full Node instance (setting the node_id and the anonymous flag). Changes to the Node message are not reflected in GetNodes and it is unclear how additional Node properties would be communicated to the driver script.

Related issues/PRs

None

Proposal

Explanation

If GetNodes returned full Node instances, one could just sample from that list and set one of the Node instances as the consumer of the new Task. Changes to the Node message (e.g., additional properties) would automatically be reflected in GetNodes.

This PR returns full Node instances instead of int node IDs from GetNodes.

Checklist

• Implement proposed change
• Write tests
• Update documentation
• Update changelog
• Make CI checks pass
• Ping maintainers on Slack (channel #contributions)

Any other comments?

[tanertopal]

@danieljanes, my thoughts on this:

• Changing from int64 to a sub-message will increase the response size, making it less scalable.
• The Driver potentially has to pull all node ids, e.g. 10M+, which was our initial thinking for choosing int64.
• If we need a more verbose response, how about splitting it into two methods?