[MISC] Flatten document encode for UI

[blythed]

Instead of nesting content inside a document, all content should get saved to the _leaves key-word field. This is a list, and may be referred to with the id field.

An example:

{
  "_leaves": [
    {
      "leaf_type": "component",
      "type_id": "datatype",
      "cls": "vector",
      "module": "superduperdb.components.vector_index",
      "dict": {
        "shape": 384,
        "identifier": "test"
      },
      "id": "_component/datatype/test"
    },
    {
      "leaf_type": "component",
      "type_id": "model",
      "cls": "SentenceTransformer",
      "module": "superduperdb.ext.sentence_transformers.model",
      "id": "_component/model/my-model",
      "dict": {
        "identifier": "my-model",
        "predict_kwargs": {
          "show_progress_bar": true
        },
        "signature": "*args,**kwargs",
        "model": "all-MiniLM-L6-v2",
        "device": "cpu",
        "postprocess": "from superduperdb import code\n\n@code\ndef my_code(x):\n    return x.tolist()\n",
        "datatype": "_component/datatype/vec384"
      }
    },
    {
      "leaf_type": "component",
      "type_id": "listener",
      "cls": "Listener",
      "module": "superduperdb.components.listener",
      "id": "_component/listener/my-listener",
      "dict": {
        "identifier": "my-listener",
        "key": "txt",
        "model": "_component/model/my-model",
        "select": {
          "documents": [],
          "query": "docu.find()"
        },
        "active": true,
        "predict_kwargs": {}
      }
    },
    {
      "leaf_type": "component",
      "type_id": "vector-index",
      "cls": "VectorIndex",
      "module": "superduperdb.components.vector_index",
      "dict": {
        "identifier": "test",
        "indexing_listener": "my-listener",
        "measure": "cosine"
      },
      "id": "_component/vector_index/test"
    }
  ],
  "my_key": "_component/vector_index/test"
}