Use presence of convention-specific sections during docstring inference

[charliermarsh]

Previously, if you had a docstring like:

def get_gas_price(data: EasyEnergyData, hours: int) -> float | None:
    """Get the gas price for a given hour.

    Args:
        data: The data object.
        hours: The number of hours to add to the current time.

    Returns:
        The gas market price value.
    """

We'd first check for any NumPy-style sections. Returns is a valid section header for both NumPy- and Google-style docstrings, so we'd infer that this is a NumPy-style docstring, and validate it as such. However, Args is a Google-style header, and isn't part of the NumPy convention.

This PR thus improves the inference logic by treating docstrings that contain Args or Arguments as Google-style, and docstrings that contain Parameters or Other Parameters as NumPy-style. If none of those sections exist, we then user the convention with the most matches (there could be weird and subtle cases here in which users are unintentionally mixing header conventions by accident, but this seems like a strict improvement).

Closes #3321.

[epenet]

Thanks @charliermarsh 👍 🥳