In a variety of cases, errors in PySeries::scatter mean Python Series.__setitem__ will corrupt the Series

[itamarst]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

s1 = pl.Series([1, 2, 3])
s1[-200] = 3  # will corrupt the series

s2 = pl.Series([1, 2, 3])
s2[[1, None] = 3  # will corrupt the series

s3 = pl.Series([object()], dtype=pl.Object)
s3[0] = 1  # will corrupt the series

Log output

No response

Issue description

The error handling path in PySeries::scatter() does not reset self.series from an empty state, so the PySeries loses all its data if you hit one of the error paths. This is a problem for exploratory data analysis, or any sort of interactive usage.

This is a spinoff from #15294

Expected behavior

Restore data to previous state, unchanged.

Installed versions

Git as of June 12, 2024.