Try out GlobalSfMPy's rotation averaging in GTSFM

[ayushbaid]

We want to try out the averaging algorithm used by the CVPR paper Revisiting Rotation Averaging: Uncertainties and Robust Losses and benchmark the results on our datasets.

This was also suggested by @sarlinpe in #754

[ayushbaid]

This is the top level python script: https://github.com/zhangganlin/GlobalSfMpy/blob/main/scripts/sfm_pipeline.py