[BUG] zfill needs to align with python standard library in some cases

[galipremsagar]

Describe the bug
Similar to pandas-dev/pandas#20868

When the string column has data which has +/- prefix signs, they will aswell need to be moved during the zfill operation.

Steps/Code to reproduce bug

In [9]: import cudf

In [10]: s = cudf.Series(["-10", "+100", "+abc", "-/bcd"])

In [11]: s.str.zfill(10)
Out[11]: 
0    0000000-10
1    000000+100
2    000000+abc
3    00000-/bcd
dtype: object

In [12]: "-10".zfill(10)
Out[12]: '-000000010'

In [13]: "+100".zfill(10)
Out[13]: '+000000100'

In [14]: "+abc".zfill(10)
Out[14]: '+000000abc'

In [15]: "-/bcd".zfill(10)
Out[15]: '-00000/bcd'

In [16]: "a+b".zfill(10)
Out[16]: '0000000a+b'

In [17]: "100-".zfill(10)
Out[17]: '000000100-'

Expected behavior
Match python standard library behavior shown above for some special cases.

Environment overview (please complete the following information)

• Environment location: [Bare-metal]
• Method of cuDF install: [from source]

[galipremsagar]

cc: @davidwendt for visibility

[davidwendt]

This was implemented to match Pandas

>>> import pandas as pd
>>> s = pd.Series(["-10", "+100", "+abc", "-/bcd"])
>>> s.str.zfill(10)
0    0000000-10
1    000000+100
2    000000+abc
3    00000-/bcd
dtype: object

Certainly seems the Python zfill is more correct.
@galipremsagar Let me know what is preferred for cuDF.

[galipremsagar]

This was implemented to match Pandas

>>> import pandas as pd

>>> s = pd.Series(["-10", "+100", "+abc", "-/bcd"])

>>> s.str.zfill(10)

0    0000000-10

1    000000+100

2    000000+abc

3    00000-/bcd

dtype: object

Certainly seems the Python zfill is more correct.

@galipremsagar Let me know what is preferred for cuDF.

Yup. The current version of zfill was implemented to match pandas. But pandas fixed this issue in 1.5.0 because python standard library zfill is correct. So we aswell need to align similarly.