Convert incoming data to appropriate resolution for all GT

[rhoadesScholar]

This should coincide with a significant refactor of how TEST_CROPS are declared:

• The repo should include a test_manifest.csv that has the following columns:

1. crop_name
2. dataset
3. class label
4. voxel_size
5. translation (on nm)
6. shape (in voxels)

• Also fetch EM data from the designated test crops #55
• Prediction regions for test crops are based on the ROIs designated in this manifest
• Function to appropriately rescale/crop predictions to match the test crops' resolutions, both before and after participant submissions. Should be called by package_submission() thus allowing participants to preview the results of the conversion, before submitting the data. Should work for both OME-NGFF metadata and classic metadata options. #54
• If no voxel size information is detectable in the incoming data, assume it is the correct resolution and crop it to match the size of the GT data.

[rhoadesScholar]

Function to appropriately rescale/crop predictions to match the test crops' resolutions, both before and after participant submissions. Should be called by package_submission() thus allowing participants to preview the results of the conversion, before submitting the data. Should work for both OME-NGFF metadata and classic metadata options.

From Saalfeld:
incoming test volumes may declare their resolution, if present and different from reference resolution, scale before evaluation, spec for scaling:

• coordinates indicate the front top left voxel corner
• binary labels are scaled using bilinear interpolation (upscaling or downscaling by factors <=2)) or volume averaging (downscaling by factors > 2) followed by thresholding
• enumerated labels are scaled using nearest neighbor interpolation (upscaling) or volume max (downscaling), special implementations for integer downscaling factors appreciated, non-integer downscaling may be approximated by downscaling to nearest integer followed by nearest neighbor downscaling for the rest

Then:

• Add above to README