Improve image search UX

[ohltyler]

Images must be in binary or base64-encoded formats to be understood by popular MM embedding models. As such, we should add helper UI around producing such formats, based on raw images users can drag into the UI. And/or, provide a URL to some image.

The rough flow can be:

1. Users opens a portal for selecting an image from local filesystem, dragging directly, or providing a url
2. User selects what format they want to convert it to (can be auto-populated based on the associated model, ideally)
3. UI handles conversion and injecting into the sample document (if ingest) or query (if search)

Furthermore, URL support is targeted to be an optional input by 2.19 - see opensearch-project/ml-commons#3054. As such, we can add URL with no further conversion as an option once that is available.