new bi sample benchmark tests

[grusev]

Reference Issues/PRs

What does this implement or fix?

    Sample test benchmark for using one opensource BI CSV source.
    The logic of a test is 
        - download if parquet file does not exists source in .bz2 format
        - convert it to parquet format
        - prepare library with it containing  several symbols that are constructed based on this DF
        - for each query we want to benchmark do a pre-check that this query produces SAME result on Pandas and arcticDB
        - run the benchmark tests

Any other comments?

Checklist

Checklist for code changes...

• Have you updated the relevant docstrings, documentation and copyright notice?
• Is this contribution tested against all ArcticDB's features?
• Do all exceptions introduced raise appropriate error messages?
• Are API changes highlighted in the PR description?
• Is the PR labelled as enhancement or bug so it appears in autogenerated release notes?