feat: add method for async read bloom filter

[hengfeiyang]

Which issue does this PR close?

Impl #3851

We want to filter row_groups in Datafusion but there is no async API for reading bloom filter.

What changes are included in this PR?

Implemented a function get_row_group_column_bloom_filter for ParquetRecordBatchStreamBuilder to support reading bloom filter outside arrow.

Are there any user-facing changes?

Add an function get_row_group_column_bloom_filter in ParquetRecordBatchStreamBuilder

[tustvold]

This looks good, left some minor comments, but I think all this needs is a test

[tustvold]

There is a new bloom_filter_length that may be present and would avoid needing to guess here

[hengfeiyang]

Thanks, i checked the module bloom_filter and then updated this part.

[tustvold]

I think it would be ideal if we could avoid this extra roundtrip in the common case, by fetching enough data in the first call

[hengfeiyang]

The first call is used to parse bloom_filter_length, and the second call is used to parse bloom_filter_data, We can reduce one call if we know the bloom_filter_length, Thanks, I updated. Can you help review again?

[hengfeiyang]

@tustvold Sure, I will try to add two test cases:

1. for the parquet file has bloom_filter_length header
2. for the parquet file has no bloom_filter_length header

[hengfeiyang]

@tustvold Can i create two test parquet files and commit to https://github.com/apache/parquet-testing/ ?

[tustvold]

You could, but I don't have merge rights there so it may take some time.

A quicker option might be to use an existing file for 1., and to write a file to a Vec for 2.

[hengfeiyang]

@tustvold Thanks, i will use {testdata}/alltypes_plain.parquet as base file and generate other files.

[mapleFU]

Would you mind take a look at data_index_bloom_encoding_stats.parquet under parquet-testing repo? I think it contains a bloom filter for the first column

[tustvold]

Looks good to me, thank you