Streaming version of polars.LazyFrame.count produces a wrong result when used in aggregation

[tikhdm]

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

import polars as pl
df = pl.DataFrame({
    'g': [1] * 2000,
    'a': ['yes', None] * 1000
}).lazy()

print('Streaming', df.group_by('g').agg(pl.col('a').count()).collect(streaming=True))
print('Non streaming', df.group_by('g').agg(pl.col('a').count()).collect(streaming=False))

The result is:

Streaming shape: (1, 2)
┌─────┬──────┐
│ g   ┆ a    │
│ --- ┆ ---  │
│ i64 ┆ u32  │
╞═════╪══════╡
│ 1   ┆ 2000 │
└─────┴──────┘
Non streaming shape: (1, 2)
┌─────┬──────┐
│ g   ┆ a    │
│ --- ┆ ---  │
│ i64 ┆ u32  │
╞═════╪══════╡
│ 1   ┆ 1000 │
└─────┴──────┘

Log output

RUN STREAMING PIPELINE
df -> primitive_group_by -> ordered_sink
RefCell { value: [] }
keys/aggregates are not partitionable: running default HASH AGGREGATION

Issue description

The behavior of polars.count is different between streaming and non-streaming queries. In the non-streaming mode it ignores empty values which corresponds to the documentation.
In the streaming mode it just returns a number of rows in each group.

Expected behavior

I would expect the same result in both streaming and non-streaming modes.

Installed versions

--------Version info---------
Polars:               0.20.15
Index type:           UInt32
Platform:             macOS-14.3.1-arm64-arm-64bit
Python:               3.12.2 (main, Feb  6 2024, 20:19:44) [Clang 15.0.0 (clang-1500.1.0.2.5)]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          <not installed>
connectorx:           <not installed>
deltalake:            <not installed>
fastexcel:            <not installed>
fsspec:               2024.2.0
gevent:               <not installed>
hvplot:               <not installed>
matplotlib:           <not installed>
numpy:                1.26.4
openpyxl:             <not installed>
pandas:               2.2.1
pyarrow:              15.0.1
pydantic:             <not installed>
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           <not installed>
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>