provide insights for insert queries

Currently query insights mostly cover select queries and targets optimizing schema to improve performance. And key contention to reduce locking delays.

At the same time we also have different type of requests where clients try to optimize raw writes and aim to achieve better throughput.

For example we document some anti-patterns like writing transactional batches could be slower than writing single rows that can use 1PC fast paths on KV level. Or bulk delete operations that generate alot of intents if applied across lots of data.

It would be nice to detect anti-patterns proactively and suggest write query optimizations as well.

Jira issue: CRDB-31870