Fix float conversion to handle blank and non string values

[Irozuku]

Summary

Uploading a CSV whose Float typed column contains blank entries (a single space, " ") crashed dataset creation with TypeError: 'NoneType' object is not iterable. comma_float_to_float caught every conversion error, printed it, and fell off the end returning None; that None was then handed to pa.table() inside transform_dataset_with_schema. Integer columns typed as Float hit the same path, since the .str accessor fails on non string series.

The conversion now treats empty and whitespace only entries as nulls, casts non string numeric arrays directly, and raises a ValueError naming the offending values instead of silently returning None.

---

Type of Change

• Backend change
• Frontend change
• CI / Workflow change
• Build / Packaging change
• Bug fix
• Documentation

---

Changes (by file)

• DashAI/back/types/utils.py: rewrote comma_float_to_float. Floating arrays still return unchanged; non string arrays (integer, decimal) cast to float64 directly; string arrays are stripped, , normalized to ., empty and whitespace only values masked to null, then converted with pd.to_numeric. Values that still fail conversion raise ValueError listing up to three samples, so the job surfaces a readable message instead of a NoneType crash. Added NumPy style docstring.

---

Testing

Upload a CSV with a Float column containing blanks. Before: job fails with Error loading dataset: 'NoneType' object is not iterable. After: blanks land as nulls, numeric values preserved as double.