Vision Service: Add support for data URLs so that no HTTPS file server is required

[lastzero]

As a user, I want to be able to add captions to my images by sending them to the service as data URLs, so that it also works behind a firewall or from the command line.

In addition, the images for which captions (or later other data) are to be generated should be read from an images array rather than from a single url field in the request, for example:

{
  "id": "d2684be3-1a80-4137-980e-07a2cbe2a96f",
  "model": "",
  "images": [
    "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAEsAAAA8CAAAAAALAhhPAAAFfUlEQVRYw62XeWwUVRzHf2"
  ]
}

See the Vision API request and unit tests in our main repository for reference:

• https://github.com/photoprism/photoprism/blob/develop/internal/ai/vision/api_request.go

Related Issues:

• AI: API to connect to external services for Image tagging / object recognition photoprism#1090 (comment)
• Labels: Load and use a custom TensorFlow model photoprism#127