This PR allows developers to define a single pydantic type which is then used in 3 key ways:

1. These objects can be constructed in Python, published to weave, and "gotten" from weave - resulting in the correct type - not a WeaveObject wrapper thing.
2. Construction of these objects can be done over the network by setting the set_base_object_class field on the POST request which will enforce server-side validation of these object payloads.
3. The frontend now has fully typed & validated read/write hooks that allows users to build UI experiences that safely read, create, and edit such objects with type safety.

A top level make command make synchronize-base-object-schemas ensures the frontend is up to date!

With all this infrustructure, you simply define your types once, like:

weave/trace_server/interface/base_object_classes/test_only_example.py:

from pydantic import BaseModel

from weave.trace_server.interface.base_object_classes import base_object_def


class TestOnlyNestedBaseModel(BaseModel):
    a: int


class TestOnlyNestedBaseObject(base_object_def.BaseObject):
    b: int


class TestOnlyExample(base_object_def.BaseObject):
    primitive: int
    nested_base_model: TestOnlyNestedBaseModel
    nested_base_object: base_object_def.RefStr

__all__ = ["TestOnlyExample", "TestOnlyNestedBaseObject"]

And you now have:

1. The ability to publish

weave.publish(TestOnlyExample(...))

2. The ability to fetch:

ref = weave.publish(TestOnlyExample(...))
assert isinstance(ref.get(), TestOnlyExample)

3. The ability to create via HTTP

curl -X 'POST' \
  'https://trace.wandb.ai/obj/create' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "obj": {
    "project_id": "user/project",
    "object_id": "my_thing",
    "val": {...},
    "set_base_object_class": "TestOnlyExample" # forces `val` to conform to the schema
  }

4. The ability to query via HTTP

curl -X 'POST' \
  'https://trace.wandb.ai/objs/query' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "project_id": "user/project",
  "filter": {
    "latest_only": true,
    "base_object_classes": ["TestOnlyExample"]
  },
}'

5. Typesafe React Hooks (Read-side)

const res = useBaseObjectInstances("TestOnlyExample", ...)
// res is typed strictly according to the pydantic schema

6. Typesafe React Hooks (Write-side)

const createFn = useCreateBaseObjectInstance("TestOnlyExample")
createFn({...})
// createFn is strictly typed and validates client-side before sending to the server!

See dev_docs/BaseObjectClasses.md for a more complete discussion