Schema parsing to-dos

[dilithjay]

• Support Gemini models: Refactor the parse_with_gemini function such that the API call goes through create_response
• Create tests for the functionality. Can create a separate file (e.g. tests/test_schema_parser.py)
• Verify and fix errors (likely with using max_tokens) when using o-series models from OpenAI
• Check if there are any benchmarks for a similar functionality. If not, start building a dataset for evaluation.
• Add support for dataclass objects, potentially with additional attributes for sample_values (examples of possible values) and alternate_keys (examples of alternative names for each attribute to support robustness across unstandardized documents).

We can create separate issues for any of the above if needed.

cc: @Vaishnav2804

[Vaishnav2804]

@dilithjay

✅ Support Gemini models: Refactor the parse_with_gemini function such that the API call goes through create_response
via PR: #108