fill_null doesn't work after using numpy ufunc

[alexanderpils]

Polars version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Polars.

Issue description

After using a numpy functions on an expression you can't use fill_null because of a SchemaError despite matching dtypes. You can make it work if you cast the expression to the same dtype again.

Reproducible example

import polars as pl
import numpy as np

df = pl.DataFrame({"a": [None,2,3,4,5,]})

df.with_columns(np.exp(pl.col("a")).fill_null(pl.lit(1, pl.Float64)))


Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/.local/lib/python3.10/site-packages/polars/internals/dataframe/frame.py", line 6022, in with_columns
    self.lazy()
  File "/home/.local/lib/python3.10/site-packages/polars/internals/lazyframe/frame.py", line 1148, in collect
    return pli.wrap_df(ldf.collect())
exceptions.SchemaError: Cannot 'fill_null' a 'Series' of dtype: 'f64' with an argument of dtype: 'f64'

Expected behavior

df.with_columns(np.exp(pl.col("a")).fill_null(pl.lit(1, pl.Float64)))

should not return an error and result in the same as

df.with_columns(np.exp(pl.col("a")).cast(pl.Float64).fill_null(pl.lit(1, pl.Float64)))

shape: (5, 1)
┌────────────┐
│ a          │
│ ---        │
│ f64        │
╞════════════╡
│ 1.0        │
│ 7.389056   │
│ 20.085537  │
│ 54.59815   │
│ 148.413159 │
└────────────┘

Installed versions

---Version info---
Polars: 0.16.7
Index type: UInt32
Platform: Linux-5.19.0-32-generic-x86_64-with-glibc2.35
Python: 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
---Optional dependencies---
pyarrow: <not installed>
pandas: <not installed>
numpy: 1.24.2
fsspec: <not installed>
connectorx: <not installed>
xlsx2csv: <not installed>
deltalake: <not installed>
matplotlib: <not installed>