More user feedback and progress bars

I find the utility of these scripts to be wonderful, but I often wonder if anything is happening. I get messages after I start processing and then it just sits there for 10, 20, 30 minutes or more. The job often does finish, but it would be nice to know if something is still happening or if it froze. If there are multiple steps, update after each step. Doing 100 iterations over 10 pictures should at least show 100 updates, if not 1000 updates (progress bar style). The more feedback the better. The same is true when loading various files. I get a message that a file is loading, but then nothing for 20 minutes. When the file is done loading, print "done loading" and then print a message for the next step. There are some messages to this effect, such as loading the BLIP file, but this can take forever and it would be nice to have some intermediate updates. For example:
```
2023-04-13 23:04:46 loading BLIP caption: https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_large_caption.pth
Downloading (…)solve/main/vocab.txt: 100%|██████████| 232k/232k [00:00<00:00, 365kB/s]
Downloading (…)okenizer_config.json: 100%|██████████| 28.0/28.0 [00:00<00:00, 22.1kB/s]
Downloading (…)lve/main/config.json: 100%|██████████| 570/570 [00:00<00:00, 317kB/s]
94%|█████████▍| 1.572023-04-14T03:50:29.282797879Z load checkpoint from https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_large_caption.pth
2023-04-13 23:50:29 BLIP loaded
```

Notice there is a 45 minute gap between timestamps. That seems a little long to not update the user that things are proceeding as expected.

Start by tracing the processing and file-loading entry points, including the BLIP caption loading path described in the issue. Identify where long-running iterations and loading steps currently emit messages; done means users receive intermediate progress updates and completion messages during processing and file loads.

Written by the indexing model from the issue text.