Improve null handling for Gemini

[dmontagu]

Closes #603

I confirmed it works and can return None by using the following script, slightly modified from the one in the issue above:

import os
from pydantic import BaseModel
from pydantic_ai import Agent
from pydantic_ai.models.gemini import GeminiModel

os.environ['GEMINI_API_KEY'] = '<scrubbed>'

class AgeModel(BaseModel):
    age: int | None = None

gemini_model = GeminiModel('gemini-1.5-flash')

prompt = "The man's age was unknown. what is the age of the old man?"
agent = Agent(gemini_model, result_type=AgeModel)
age = agent.run_sync(prompt)
print(age.data)
# age=None

@samuelcolvin let me know if you want any testing beyond the one I updated here.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[dmontagu]

I'm not sure why we are currently requiring the default to be None in order to remove the the {'type': 'null'} from the schema, that condition doesn't seem to me to be necessary, so I removed it here. (And we don't use the value of the default at all, so I removed the variable that was holding its value, and changed the .pop call to fall back to None to be consistent with the previous line which is also unused.)

[samuelcolvin]

thank you.