[Vertex AI] Add ImageGenerationInstance for input to predict call

[andrewheard]

Added an encodable type ImageGenerationInstance that is passed in instances[] in a predict request. The schema (download) supports other fields, such as image and mask, but we are only using prompt at this time.

VisionGenerativeModel Instance Schema Reference

title: VisionGenerativeModel
type: object
required:
- prompt
properties:
  prompt:
    type: string
    description: >
      The text prompt for guiding how vision model generate the images.
      This field is required for both generation and editing.
  image:
    type: object
    description: >
      Image for editing. This field is required for editing. Not needed for generation.
    oneof:
    - bytesBase64Encoded
    - gcsUri
    properties:
      bytesBase64Encoded:
        type: string
        description: Image bytes encoded in base64 string.
      gcsUri:
        type: string
        description: >
          The Google Cloud Storage location of the image on which to perform the editing.
        pattern: '^gs:\/\/(.+)\/(.+)$'
      mimeType:
        type: string
        description: >
          The MIME type of the content of the image. Only the images in below listed
          MIME types are supported.
        enum:
        - image/jpeg
        - image/png
  mask:
    type: object
    description: >
      Mask where to edit. This is an optional field for editing. No need for generation.
    oneof:
    - image
    - polygonList
    properties:
      polygonList:
        type: array
        description: Multiple polygon masks.
        items:
          type: array
          description: >
            All of the vertices that form the single polygon mask.
          items:
            type: object
            description: >
              Single vertex with `x` and `y` coordinates that describes a point in 2D plane.
            properties:
              x:
                type: number
                format: float
                minimum: 0.0
                maximum: 1.0
                description: The x coordinate of the vertex.
              y:
                type: number
                format: float
                minimum: 0.0
                maximum: 1.0
                description: The y coordinate of the vertex.

#no-changelog