chore: workaround exponential planning time bug in DF fork

[alamb]

Targets https://github.com/influxdata/influxdb_iox/pull/13448

As @wiedld mentions in #54 (comment):

The upstream code has been changed so our previous revert can't be applied cleanly

This PR reverts the behavior with a different patch (copy/paste the old code from apache#12562)

Locally when I run the following command without this fix, it fails (timesout)

$ RUST_BACKTRACE=1 TEST_INFLUXDB_JDBC=true TEST_INTEGRATION=1 TEST_INFLUXDB_IOX_CATALOG_DSN=postgresql://postgres@localhost:5432/postgres cargo test --test  end_to_end --  read_filter_many_tables

When I have this fix it passes

running 1 test
test end_to_end_cases::querier::influxrpc::read_filter::read_filter_many_tables ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 478 filtered out; finished in 9.67s

[alamb]

These are tests of the reverted behavior so I removed them in this commit

[alamb]

I expect clippy to fail due to the arrow 54.1 release -- I expect it till be fixed with the actual datafusion 45 release (as I will port this PR):

• chore: Upgrade to arrow/parquet 54.1.0 and fix clippy/ci apache/datafusion#14415

[alamb]

This was included in #54