Default DataFrame source to pandas driver when none is given (fixes #1403)

[gaoflow]

Problem

Defining a DataFrame source without an explicit driver raised a ValidationError (#1403):

data_catalog = hydromt.DataCatalog()
data_catalog.from_dict({
    "agroforestry_dataframe": {
        "data_type": "DataFrame",
        "uri": "path/to/dataframe.csv",
        # no "driver" specified
    },
})
# pydantic_core.ValidationError: ... driver
#   Value error, Unknown 'name': 'geodataframe_table'

DataSource._infer_default_driver searches all drivers (BaseDriver.find_all_possible_types()) for one that supports the file extension. Both the DataFrame pandas driver and the GeoDataFrame geodataframe_table driver claim .csv/.parquet, and geodataframe_table is enumerated first — so it gets selected, which is not a valid driver for DataFrameSource.

Fix

Override _infer_default_driver on DataFrameSource to restrict the search to DataFrameDriver subclasses, mirroring the existing override on GeoDatasetSource. RasterDatasetSource, GeoDataFrameSource and GeoDatasetSource already do this; DatasetSource is unaffected because no other driver claims its extensions.

Reproduction / verification

import hydromt
dc = hydromt.DataCatalog()
dc.from_dict({"df": {"data_type": "DataFrame", "uri": "path/to/dataframe.csv"}})
print(dc.get_source("df").driver.name)
# before: ValidationError (geodataframe_table)
# after : "pandas"

• New tests in tests/data_catalog/sources/test_dataframe_source.py: _infer_default_driver returns "pandas" for .csv/.parquet/.xlsx, and model_validate of a driver-less DataFrame dict yields the pandas driver. They fail on main and pass with this change. An explicitly specified driver is still honoured.
• Full tests/data_catalog/sources/ suite passes (81 passed); the broader tests/data_catalog/ run is green apart from one pre-existing failure that requires the Deltares P-drive (test_export_deltares_data[grwl_mask]), unrelated to this change.
• ruff check and ruff format --check clean.

[gaoflow]

I checked the current red CI jobs for this branch.

• pytest low-314 (ubuntu-latest) has a single failure in tests/data_catalog/test_data_catalog.py::test_azure_get_rasterdataset[esa_worldcover_azure]. The traceback is from fetching the Azure SAS token and ends with ConnectionResetError: [Errno 104] Connection reset by peer, so this looks like an external Azure/network integration failure rather than the DataFrame driver change.
• run test coverage and SonarQube scan completed the test suite successfully (1018 passed, 9 skipped, 77 deselected) and then failed in the SonarQube upload because the workflow does not have an authorized SONAR_TOKEN in this context.

I did not push code changes because both failures are outside the scope of this PR.