Create a metric to track the number of queries with an index join over time

It's difficult to quickly determine whether workload performance is tuned over time. Customer workloads can easily change and/or table schema can be altered which can cause regressions to creep and degrade performance as plans shift requiring an index join. Index joins can indicate a suboptimal indexing strategy. This issue tracks creating a high-level metric similar to our full table scan metric for index joins.

Jira issue: CRDB-30103