Support separate null_values per column in pyarrow.csv.ConvertOptions

[paiforsyth]

Describe the enhancement requested

I have a csv dataset in which nulls are encoded differently in different columns. It looks like when reading csv data with pyarrow, the same list of null_values must be used for all columns (see ConvertOptions). This concerns me because a value used as a null code in one column ("9999" for example) may be a valid non-null value in another column. In pandas's read_csv, it is possible to pass a dictionary specifying different null codes for different columns. Could this functionality be added to pyarrow?

Component(s)

Other