[BUG] pack and unpack causing wrong behaviour with slices of struct columns.

[shaneding]

Describe the bug
After fixing slicing from a struct column in #8719, tests inside test_pack.py are now failing due to pack and unpack not correctly dealing with struct column slices. This is likely an issue with offset (same reason as what was causing the old error with slicing).

Steps/Code to reproduce bug

In [37]: import cudf

In [38]: df = cudf.DataFrame()

In [39]: df['a'] = cudf.Series([{'a': i} for i in range(5)])

In [40]: df
Out[40]: 
          a
0  {'a': 0}
1  {'a': 1}
2  {'a': 2}
3  {'a': 3}
4  {'a': 4}

In [41]: df[2:]
Out[41]: 
          a
2  {'a': 2}
3  {'a': 3}
4  {'a': 4}

In [42]: from cudf._lib.copying import pack, unpack

In [43]: packed = pack(df[2:])

In [44]: unpacked = cudf.DataFrame._from_table(unpack(packed))

In [45]: unpacked
Out[45]: 
          a
2  {'a': 0}
3  {'a': 1}
4  {'a': 2}

Expected behavior
Line 45 should output the same as line 41

Additional context
In the current version of cudf, line 41 will output the same result as line 45, which will cause this test to pass, however that is incorrect behaviour as the slice is not correct. #8719 will fix this slicing issue, however it will not pass the CI as a result of this test case. We may need to ignore test_pack.py until #8719 is merged then address this pack and unpack issue in a later PR, or address it altogether in #8719.

[quasiben]

@charlesbluca is this something you can look at ?

[shaneding]

@quasiben This has been resolved via #8739. I will close the issue.

[quasiben]

Ah, great! I've been slowly going through the my backlog and came across this one -- I'll see your PR in a few more clicks