Unify temporal expression functions

[romanovacca]

Problem description

A couple functions are called similar, while having different behaviour e.g. Expr.dt.microsecond() and Expr.dt.microseconds().

These functionality of these could be unified. See #6413 for more information

[MarcoGorelli]

The current names were (probably?) taken from pandas / Python stdlib, but they behave differently:

In [31]: pd.Series([timedelta(days=1, hours=1)]).dt.seconds
Out[31]:
0    3600
dtype: int32

In [32]: pl.Series([timedelta(days=1, hours=1)]).dt.seconds()
Out[32]:
shape: (1,)
Series: '' [i64]
[
        90000
]

The pandas (and Python stdlib) equivalent of Polars .dt.seconds is actually .dt.total_seconds:

In [41]: pd.Series([timedelta(days=1, hours=1)]).dt.total_seconds()
Out[41]:
0    90000.0
dtype: float64

The Python standard library behaves in the same way:

In [49]: timedelta(days=1, hours=1).seconds
Out[49]: 3600

In [50]: timedelta(days=1, hours=1).total_seconds()
Out[50]: 90000.0

So, rather than unifiying temporal .dt.microseconds and datetime .dt.microsecond, maybe all that needs doing is:

• .dt.seconds => .dt.total_seconds
• .dt.microseconds => .dt.total_microseconds
• ...
• .dt.nanoseconds => .dt.total_nanoseconds
  ?

This should

• reduce the surprise for users used to pandas / Python stdlib
• reduce the risk of accidentally confusing them because of a single character difference (this is quite easily done, I've done it too 😄 )