Value model in AdditionalFieldDefinition is an empty class, thus silently drops all data

The Value class in additionalfielddefinition.py is defined as an empty Pydantic BaseModel with no fields:

class Value(BaseModel):
  r"""Either a string or HypertextField. When OpenAPI Generator supports oneOf, we can semantically enforce this in the docs."""

Because the SDK's BaseModel uses the default Pydantic extra = 'ignore' behavior, any data passed to Value(...) is silently discarded. This means the value field on AdditionalFieldDefinition, which is typed as Optional[List[Value]], cannot actually carry any data.
Impact

This affects any consumer using AdditionalFieldDefinition through either:

• EmployeeInfoDefinition.additional_fields
• TeamInfoDefinition.additional_fields

When constructing an AdditionalFieldDefinition with values, the data in the value list is lost:

from glean.api_client.models.additionalfielddefinition import AdditionalFieldDefinition, Value
field = AdditionalFieldDefinition(
  key="languages",
  value=[Value(anchor="English"), Value(anchor="Spanish")]
)
print(field.model_dump())
# {'key': 'languages', 'value': [{}, {}]}
#                                 ^^   ^^  all data silently dropped

The same applies to HypertextField-style values:

Value(anchor="Glean", hyperlink="https://glean.com")
# Value()  — both anchor and hyperlink are silently ignored

This is the associated public docs page: https://developers.glean.com/api/indexing-api/bulk-index-employees. Note that even when passing in data to Value() as a dict, the whole dict object is ignored:

from glean.api_client.models.additionalfielddefinition import AdditionalFieldDefinition

raw = {"key": "username", "value": [{"username": "testuser"}]}
field = AdditionalFieldDefinition(**raw)
dumped = field.model_dump()
print(dumped)
# {'key': 'username', 'value': [{}]}

Start in glean/api_client/models/additionalfielddefinition.py by inspecting the empty Value model and how AdditionalFieldDefinition constructs its value list. Reproduce the documented examples with anchor, hyperlink, and dictionary inputs, then confirm that model_dump() preserves the supplied data instead of producing empty objects.