[BUG] Assigning incompatible NA types to a column is leading to implicit conversions

[galipremsagar]

Describe the bug
When we try assigning cudf.NaT to non datetime/timedelta types, we are implicitly type-casting to cudf.NA. Pandas actually assign object dtype or raises incase of nullable dtypes.

Steps/Code to reproduce bug

In [19]: import cudf

In [20]: import pandas as pd

In [21]: s = pd.Series([1, 2, 3])

In [22]: s
Out[22]: 
0    1
1    2
2    3
dtype: int64

In [26]: s[1] = pd.NaT

In [27]: s
Out[27]: 
0      1
1    NaT
2      3
dtype: object

In [28]: s = cudf.Series([1, 2, 3])

In [29]: s
Out[29]: 
0    1
1    2
2    3
dtype: int64

In [30]: s[1] = cudf.NaT

In [31]: s
Out[31]: 
0       1
1    <NA>
2       3
dtype: int64

Expected behavior
Raise an error.

Environment overview (please complete the following information)

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

[mroeschke]

In 3.0, In [26]: will be disallowed (the user will need to cast the Series to the correct scalar type before assigning to avoid object)