[BUG] Cannot read non-annotated fixed_len_byte_array data from Parquet as binary (string)

[NVnavkumar]

Describe the bug
Parquet supports the FIXED_LEN_BYTE_ARRAY physical type in addition to the BINARY physical type (which is basically a variable length byte array). cuDF can currently read Decimals and other types which are stored as FIXED_LEN_BYTE_ARRAY, but can't read the FIXED_LEN_BYTE_ARRAY data as binary data (or string). It should behave the same as BINARY.

Basically, this schema:

message table {
| required FIXED_LEN_BYTE_ARRAY(10) bin_test;
|}

should be handled the same as:

message table {
|    required BINARY bin_test;
|}

This is required for NVIDIA/spark-rapids#7449

Steps/Code to reproduce bug
flba_binary_parquet.zip

The attached parquet file has one row with the following schema:

message spark {
            |  required fixed_len_byte_array(5) a;
            |  required fixed_len_byte_array(5) b;
            |}

In cuDF Python:

>>> df = cudf.read_parquet("/tmp/flba_binary.parquet")
>>> df
  a b
0

Expected behavior

The output should be:

>>> df
       a      b
0  hello  there

Environment overview (please complete the following information)

• Docker commands:

docker pull rapidsai/rapidsai-core-nightly:23.06-cuda11.8-runtime-ubuntu22.04-py3.10
docker run --gpus all --rm -it \
        --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 \
        -p 8888:8888 -p 8787:8787 -p 8786:8786 \
        rapidsai/rapidsai-core-nightly:23.06-cuda11.8-runtime-ubuntu22.04-py3.10

Ran ipython inside Docker container

[revans2]

I think this is a duplicate of #12590

[NVnavkumar]

Closing this as duplicate of #12590