fix: handle numeric values in JSON schema processing

[yashwantbezawada]

Fixes #279

Problem

The rec_strict_json_schema function raises a ValueError when processing Pydantic models with constrained types like conlist, conset, etc.

These constrained types generate JSON schemas with numeric constraint values (e.g., minItems: 3, maxItems: 5), which are valid JSON schema properties. However, the function only handled strings, booleans, dicts, lists, and None, causing it to fail on these numeric values.

Changes

Updated rec_strict_json_schema to accept int and float as valid primitive types, similar to how it handles str and bool. These values are passed through unchanged during schema processing.

Also updated the test suite to verify that numeric values are handled correctly and added a specific test case for conlist models.

Testing

Verified with the exact reproduction case from the issue:

from pydantic import BaseModel, conlist
from mistralai.extra.utils.response_format import response_format_from_pydantic_model

class ExampleModel(BaseModel):
    words: conlist(str, min_length=3, max_length=3)

response_format = response_format_from_pydantic_model(ExampleModel)

This now works without raising a ValueError.