[BUG] cudf should throw exception when reading parquet file with unsupported codec

[wbo4958]

Describe the bug
Cudf returns wrong values when reading parquet file with zstd codec in the latested 22.04 nightly build. The file is test.zstd.txt.

Or Cudf throws exception rmm-src/include/rmm/cuda_stream_view.hpp:81: cudaErrorIllegalAddress an illegal memory access was encountered, which I did not reproduce.

Meta info of test.zstd.txt

creator:     parquet-mr version 1.12.1 (build 2a5c06c58fa987f85aa22170be14d927d5ff6e7d) 
extra:       org.apache.spark.version = 3.2.0 
extra:       org.apache.spark.sql.parquet.row.metadata = {"type":"struct","fields":[{"name":"a","type":"integer","nullable":false,"metadata":{}}]} 

file schema: spark_schema 
--------------------------------------------------------------------------------
a:           REQUIRED INT32 R:0 D:0

row group 1: RC:3 TS:35 OFFSET:4 
--------------------------------------------------------------------------------
a:            INT32 ZSTD DO:0 FPO:4 SZ:44/35/0.80 VC:3 ENC:BIT_PACKED,PLAIN ST:[min: 1, max: 3, num_nulls: 0]

cudf read

In [1]: import cudf

In [2]: df = cudf.read_parquet("parquet-zstd")

In [3]: df.a
Out[3]: 
0    768
1    768
2      0
Name: a, dtype: int32

pandas read

In [14]: import pandas as pd

In [15]: df = pd.read_parquet("parquet-zstd")

In [16]: df.a
Out[16]: 
0    1
1    2
2    3
Name: a, dtype: int32

Expected behavior

Although Cudf does not support zstd compression, it should throw an exception with "unsupported zstd codec" instead of returning the wrong value or cudaErrorIllegalAddress ...