Add an endpoint to recheck an entire folder for large images

[manthey]

This would check if any files could be used as large images that aren't marked as such, and would check if existing large_images are still readable, and, if not, unmark and try to remark them. It should have a recurse option.

Ideally, this would be exposed as a local job so that it can be cancelled.

[manthey]

This is related to #873.

[manthey]

I think the basic process would be to iterate through all the items in a folder or folder tree. Pseudocode:

from girder_large_image.models.image_item import ImageItem

for item in Folder().childItems():
    if item.get('largeImage'):
        if item['largeImage']['expected']:
            # we asked to create a large image and it hasn't finished.  Either the process is in the works or failed;
            # maybe a flag on the endpoint would optionally `ImageItem().delete(item)` in this case and then fall through
            # to the condition is largeImage isn't present in item.
            pass
        else:
            try:
                ItemItem().getMetadata(item)
                # all well, continue
                continue
            except Exception:
                 # we failed to open the item and get its image metadata, therefore:
                 previousFileId = item['largeImage'].get('originalId', item['largeImage']['fileId'])
                 ImageItem().delete(item)
                 # also allow creating jobs based on an option
                 ImageItem().createImageItem(item, File().load(previousFileId))
    else:
        # item was never a large image, can we ask to make sure it can't be?
        # get file id the same way as the createTiles method in the girder large_image rest endpoint
        # also allow creating jobs based on an option
        ImageItem().createImageItem(item, File().load(previousFileId))

Options we could want for this endpoint (this can be multiple PRs):

• recurse the specified folder
• delete creation jobs (default False)
• use creation jobs if we can't read files as images directly (default False)
• optional criteria to force undoing the largeImage and redoing it (the two examples I have are a regex for the item name or a source name)
• always use a job when creating the largeImage (the same action as when we ctrl-click the large image button in the UI)

[manthey]

There are a few motivations for this endpoint.:

• It is possible that the format support was added or improved since the items were uploaded
• Previous versions might have picked a poor source because of a partial upload. Example, mrxs files have a folder of data and an item that can be interpreted as a JPEG. If the item is selected as a large image before the folder is fully uploaded, we read it with the pil source as a jpeg; redoing it will read it with the openslide source as a mrxs file (and have much more detail).
• Some files may have gone offline; unmarking them reduces the attempt to use them (this is probably indicative of bad asset management, but that should be addressed differently).
• If we optionally use a job to reencode an image, it will use more space but also be faster to access. We have cases where we have a folder full of un-optimized images and reencoding all of them in one API call would be handy.