[BUG/DOC] cudf-polars does not return `precision=None` decimal dtypes

[Matt711]

Describe the bug
See the repro

Steps/Code to reproduce bug

In [5]: import polars as pl
   ...: from decimal import Decimal
   ...: 
   ...: df = pl.LazyFrame(
   ...:     {
   ...:         "bar": [Decimal("1.23"), Decimal("4.56"), Decimal("7.89")],
   ...:     },
   ...:     schema={"bar": pl.Decimal(scale=2)},
   ...: )
   ...: 
   ...: df.collect()
Out[5]: 
shape: (3, 1)
┌──────────────┐
│ bar          │
│ ---          │
│ decimal[*,2] │
╞══════════════╡
│ 1.23         │
│ 4.56         │
│ 7.89         │
└──────────────┘

In [6]: df.collect(engine="gpu")
Out[6]: 
shape: (3, 1)
┌───────────────┐
│ bar           │
│ ---           │
│ decimal[38,2] │
╞═══════════════╡
│ 1.23          │
│ 4.56          │
│ 7.89          │
└───────────────┘

Expected behavior
We can document this behavior for now. Given that polars will eventually remove precision, we should be able to match polars CPU exectly at that time.