Utilize colmap's EXIF metadata extraction feature in ns-process-data images pipeline to faciltate georegistration.

[SmolBoi96]

COLMAP by default will extract EXIF Metadata (if they exist) from images. For example focal length and GPS data will be used as priors. Nerfstudio is not utilizing this feature because it runs colmap on the copied images which does not have these EXIF data.

Suggestion:
Run colmap feature extractor on the original images with EXIF data. This will automatically extract EXIF data and populate the pose prior and focal length priors section in the database. The database will now have the original image names instead of "frame_XXXXX.jpg". To get the copied images to work on this database we can copy the original filename as well instead of using the "frame_XXXXX.jpg" name.

Benefits:

1. Pose priors allow the use of spatial matcher which matches images based on camera proximity. This is the preferred matching if position priors are known. (acoording to https://colmap.github.io/tutorial.html). I propose we enable spatial matching as an option in ns-process-data images pipeline.

2. Pose priors allow the colmap sparse data to be easily georectified. I propose we also enable a flag to do georegistration using colmap. Refer to this pipeline:
   Document how to populate the database with pose priors and use the PosePriorBundleAdjuster colmap/colmap#2976 (comment)

[ishipachev]

This will automatically extract EXIF data and populate the pose prior

But so far COLMAP doesn't support pose priors, only position priors.

I think so far it can be resolved by just running COLMAP at first and feeding its DB to Nerfstudio. In the end I believe COLMAP still have reconstruction in local normalized reference frame.