Easy-to-Share Model Checkpoints in Nerfstudio

[AntonioMacaronio]

Problem Background

• For a while, sharing nerfstudio model checkpoints was actually very painful because when creating a nerfstudio pipeline, the original nerfstudio dataset is needed in order to view the model.
• In addition to the original dataset, the receiver must also have their directory structure exactly replicate the way the person who trained the model because the nerfstudio configs use absolute filepaths.
• Ideally, a nerfstudio model (whether it's a NeRF or a 3DGS) should be decoupled from its data when sharing model checkpoints.

Overview of Changes

• Creation of a "MockDataparser" which the pipeline can use to instantiate itself