[FR] Update Labelbox integration to support loading video object tracks and immutable attributes

[ehofesmann]

Today, when loading object tracks on videos annotated through our Labelbox integration, the index field is not populated. This means that the resulting objects are not technically tracks, but frame-level detections. The Labelbox videos exports seem to support some form of unique ID per object that we could use to link together sequential objects into tracks when loading them.

Similarly, attributes that are being annotated on objects for video tasks could have the mutable attribute set to False, where the intention is that the attribute cannot change throughout the course of a given object track in the annotation tool. More investigation is needed if classification scopes could be used to set immutable attributes.