Add a node for inputting coordinates to SAM3 Detect node

### Feature Idea

SAM3 Detect node has inputs for `positive_coords` and `negative_coords`, but the only way to provide those values (that I'm aware of) is to use a text node and manually type in a JSON string with an array of coordinates. Instead, there should be a simple node that would allow the user to provide multiple sets of coordinates and output them in the JSON format that SAM3 Detect node expects.

It would also be helpful if the mask editor could display current mouse coordinates somewhere on the screen. So that we don't have to go to an external image editor to find coordinates for SAM.

### Existing Solutions

_No response_

### Other

_No response_

Start by locating the SAM3 Detect node and tracing how positive_coords and negative_coords are currently supplied from a text node. Then inspect the mask editor entry point. Done means users can provide multiple coordinate sets in the expected JSON format and can see current mouse coordinates in the editor.

Written by the indexing model from the issue text.