fix(rust, python)!: don't parse inputs without hour and minute components as pl.Datetime (use pl.Date instead)

[MarcoGorelli]

closes #8849

This makes things slightly inconvenient for anyone previously parsing '2020-01-01' as Datetime, as they need to parse Date and then cast to Datetime, but better than silently dropping the 'hour' component IMO

To clarify:

Current:

In [22]: pl.Series(['2020-01-01']).str.to_datetime('%Y-%m-%d')
Out[22]:
shape: (1,)
Series: '' [datetime[μs]]
[
        2020-01-01 00:00:00
]

Here:

In [24]: In [22]: pl.Series(['2020-01-01']).str.to_datetime('%Y-%m-%fda')
---------------------------------------------------------------------------
ComputeError: strict conversion to datetimes failed.

You might want to try:
- setting `strict=False`
- explicitly specifying a `format`
- setting `utc=True` (if you are parsing datetimes with multiple offsets)
- using `.to_date` instead of `.to_datetime`, or `.strptime(pl.Date)` instead of `.strptime(pl.Datetime)`

In [23]: pl.Series(['2020-01-01']).str.to_date('%Y-%m-%d')  # can cast to `Datetime` if they wish
Out[23]:
shape: (1,)
Series: '' [date]
[
        2020-01-01
]

[ritchie46]

@MarcoGorelli any update on this? Or shouldn't we include this in 0.18 and do we first need to iron things out?

[MarcoGorelli]

hey - just wanted to check we were aligned, are you on board with the reasoning in #8849 (comment) ? If so I'll resolve the conflicts and push

[MarcoGorelli]

superseded by #9044