cluster tree created nodes be excluded when extracting exemplars during modifying PredictionData

[Dicksonchin93]

cluster tree nodes (not nodes from the raw data) should be excluded when extracting exemplars during initializing or modifying PredictionData in https://github.com/scikit-learn-contrib/hdbscan/blob/94744a5715a639ecb084e803f96ddf6c909c3e07/hdbscan/flat.py#L789-L802 https://github.com/scikit-learn-contrib/hdbscan/blob/94744a5715a639ecb084e803f96ddf6c909c3e07/hdbscan/prediction.py#L134-L143

When cluster tree nodes are part of the child nodes that got selected to be the exemplar, this part of the code fails https://github.com/scikit-learn-contrib/hdbscan/blob/94744a5715a639ecb084e803f96ddf6c909c3e07/hdbscan/flat.py#L802 https://github.com/scikit-learn-contrib/hdbscan/blob/94744a5715a639ecb084e803f96ddf6c909c3e07/hdbscan/prediction.py#L143

Should the cluster nodes be removed before trying to extract from the raw data? Thoughts anyone?