Migrate model inference to use `doodleverse_utils`

[CameronBodine]

I spoke with @dbuscombe-usgs about the PING-Mapper inference workflows currently under development. These workflows utilize similar approaches as those developed in segmentation_gym. doodleverse_utils is the "engine" behind both segmentation_gym and doodler, both of which are used extensively in the forthcoming improvements to PING-Mapper v2.0. Therefore, doodleverse_utils should be added as a pip dependency, allowing for seamless integration of existing and future utilities.

The following updates are necessary to the dev branch:

• [X] Add doodleverse_utils as a pip dependency in PINGMapper.yml
• [X] Update automated depth detection
• [X] Update automated shadow detection
• [x] Update automated substrate detection