Misleading error message when using incorrect types for UDFs

[YLGH]

When giving an UDF an unexpected/unsupported input type, the error msg is misleading. For example

import torcharrow as ta
from torcharrow import functional as F_eager

df = ta.dataframe(
    {
        "a": [x for x in range(5)],
        "b": [100 * x for x in range(5)],
        "label": [np.float32(x + 0.5) for x in range(5)],
        "label_str": [str(np.float32(x + 0.5)) for x in range(5)],
    }

)

print(F_eager.torcharrow_round(df["label"]))
print(F_eager.torcharrow_round(df["label_str"]))

The first one prints without any issue,

the second shows a

Request for unknown Velox UDF: torcharrow_round

since it doesn't expect a Column(str)

However, my first impression was to think that this method doesn't exist. Would be great if we could return something along the lines of "torcharrow_round exists, but only accepts int/floats/etc"

[wenleix]

To start with, even put the type get called would be useful (e.g. cannot find "torcharrow_round" with input type VARCHAR)

[wenleix]

Actually, looks like we can do something similar to https://github.com/facebookincubator/velox/blob/041f3759d6e4c070c2ecd2e01eae6a7be9d93957/velox/parse/TypeResolver.cpp#L112-L124

[wenleix]

resolved via #400