join statement with regex causes panic

[PPL143]

Describe the bug

Using regex functions or regex logical operators will result to overflow.

To Reproduce

> CREATE EXTERNAL TABLE bigdata STORED AS CSV LOCATION 'bigdata.csv' OPTIONS ('has_header' 'true');
0 row(s) fetched.
Elapsed 0.007 seconds.

>
CREATE EXTERNAL TABLE find STORED AS CSV LOCATION 'find.csv' OPTIONS ('has_header' 'true');
0 row(s) fetched.
Elapsed 0.001 seconds.

>

select count(1) from bigdata;
+-----------------+
| count(Int64(1)) |
+-----------------+
| 2278791         |
+-----------------+
1 row(s) fetched.
Elapsed 0.023 seconds.

> select count(1) from find;
+-----------------+
| count(Int64(1)) |
+-----------------+
| 27              |
+-----------------+
1 row(s) fetched.
Elapsed 0.001 seconds.

>
select * from bigdata join find ON bigdata.a ~* find.regex;

thread 'tokio-runtime-worker' panicked at ..\index.crates.io-1949cf8c6b5b557f\arrow-select-55.1.0\src\take.rs:480:45:
overflow

Expected behavior

No response

Additional context

data.zip

[chenkovsky]

take

[chenkovsky]

https://github.com/apache/arrow-rs/blame/6deefb7d6163dceb360236078d5d9f10d1021422/arrow-select/src/take.rs#L480

When the length overflows, it will panic, which may not be a good idea. Because overflow does not necessarily mean there is a problem with the program. In the current case, it is just because the indices array is too large. One possible solution is to automatically convert to LargeUtf8 when we find overflow, but this may break some schema consistency checks. Another solution is to allow returning Vec. But these will have some API changes.