Let's write benchmarks for critical operations of the driver, to be aware of how fast we are and to gauge impact of further optimisations
/ detect regressions.

Proposed functionality to be benchmarked:

• INSERTs with and without data,
• SELECTs with little and a lot of data,
• both prepared and unprepared statements,
• batches,
• concurrent executions API,
• TODO - more ideas welcome.