Introduce integration-level benchmarks

[noisersup]

What should be done?

Let's introduce benchmarks for a various queries to check if we don't regress in performance.

Let's create 2 sets of tests

Beta Give feedback

1. Inserting large amount of documents and querying them
2. Inserting one large document, find and replace it Add benchmark for replacing large document #2482

• We shouldn't use any external data for that.
• All tests should compare time between 2 databases
• We should also create subsets that will run tests with and without pushdown
• We should compare benchmark results with previous one (benchstat)

Where?

• integration package. (Consider using subpackage)