In our use case, it would be extremely useful to allow specifying Feature-s for compute nodes in ComputeResource. The specified features list should be merged with the features that ParallelCluster currently generates.

Possible generic use case: Cluster admin can specify features for compute nodes and then either create custom partitions with fetaure-based filters or allow users to filter nodes for their jobs directly by feature.