Add CollectionIndexOnNonIndexedSeq rule

[JoshRosen]

Inspired by https://twitter.com/jshrsn/status/865677863646658560, this patch adds a new IndexingOnANonIndexedSeq rule which raises a warning when we see code which indexes on a Seq which isn't known to be an IndexedSeq subtype. The goal here is to make it easier to prevent cases where someone writes an O(n^2) loop by accident.

This rule might have prevented the following Spark bugs:

• [SPARK-20773][SQL] ParquetWriteSupport.writeFields is quadratic in number of fields  apache/spark#18005
• [SPARK-9952] Fix N^2 loop when DAGScheduler.getPreferredLocsInternal accesses cacheLocs apache/spark#8178
• [SPARK-17648][CORE] TaskScheduler really needs offers to be an IndexedSeq apache/spark#15221

In order to avoid false-positives from the use of .apply(0) instead of .head I chose to ignore all cases where the index is a literal.

[sksamuel]

👍