substr_index doesn't behave correctly with a negative delimiter position

[sanderson]

Describe the bug

The substr_index function only returns the expected response for the first row in a table and an incorrect result for all subsequent rows when using a negative delimiter count. It seems to use the position of the specified delimiter in the first row to process all subsequent rows. It also includes the delimiter in the returned substring, when I don't believe that it should.

To Reproduce

Run the following query:

SELECT
  url,
  substr_index(url, '.', 1) AS subdomain,
  substr_index(url, '.', -1) AS tld
FROM
  (values ('docs.apache.com'),
          ('community.influxdata.com'),
          ('arrow.apache.org')
  ) data(url)

url	subdomain	tld
docs.influxdata.com	docs	.com
community.influxdata.com	community	xdata.com
arrow.apache.org	arrow	e.org

Expected behavior

I would expect the above query to return the following:

url	subdomain	tld
docs.influxdata.com	docs	com
community.influxdata.com	community	com
arrow.apache.org	arrow	org

Additional context

I tested this in InfluxDB v3's implementation of DataFusion, but I believe these changes were just pulled from upstream.

[jonahgao]

take