Create JOB benchmark

JOB (join order benchmark) is a good benchmark with real data.

We don't care about join order here, but we want the real data with filters.

The end goal for this issue is two parts:

1. Find a way to automatically set up the data necessary for job, in parquet format.
2. Generate queries based on the queries in job.

More specifically,

1. download imdb dataset, convert it into parquet format. this should be done using a simple script.
2. To generate queries:

Given a query from job:

SELECT MIN(t.title) AS movie_title
FROM company_name AS cn,
     keyword AS k,
     movie_companies AS mc,
     movie_keyword AS mk,
     title AS t
WHERE cn.country_code ='[sm]'
  AND k.keyword ='character-name-in-title'
  AND cn.id = mc.company_id
  AND mc.movie_id = t.id
  AND t.id = mk.movie_id
  AND mk.keyword_id = k.id
  AND mc.movie_id = mk.movie_id;

We'd like to remove all unrelated queries nodes and make it into simple:

SELECT cn.country_code, k.keyword
FROM company_name AS cn,
     keyword AS k,
WHERE cn.country_code ='[sm]'
  AND k.keyword ='character-name-in-title';

Finally, we'd like to run the query and data in the ci.

Start with the linked IMDb dataset and the JOB SQL queries; identify the repository's existing approach for scripts, Parquet data, and CI. Done means an automated setup produces the required Parquet data, generates simplified filter queries from JOB, and runs the resulting queries and data in CI.