simplify workflow of upload input and label images, visualize in ng, then segment them

[samcraftt]

Before:

• upload the images in the drag and drop area
• click each to confirm their types (which one is the input, which one is the labels)
• then select from a dropdown menu your input image, and another dropdown menu for which is your label image
• click visualize
• when you go to model inference tab, would have to again select your input and labels from dropdowns

Could use some simplification/streamlining for the mito mvp demo purposes, so now the process is:

1. upload the input and label in their respective drag and drops:

2. then click visualize, you get this:

3. from this page you can just click "Segment it" which does the equivalent of clicking to the "Model Inference" tab, with the correct, previously uploaded files already selected in the dropdown:

[samcraftt]

I added a small tweak. the drag and drops were only accepting image/* file types, but this wasn't there before and it shouldn't be there, doesn't need to be there