Parameterised query benchmark

[Mike-Dax]

This PR adds a benchmark to compare the performance of parameterising a query.

The statistically significant takeaway by my reading is:

• It's faster to prepare once then re-bind parameters multiple times if possible.
• If the query can't be parameterised, calling prepare and binding every time has an overhead that can be avoided with a simpler run call.

 ✓ test/bench/prepare.bench.ts (3) 2041ms
   ✓ Parameterised queries (3) 2040ms
     name                        hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · Multiple queries      2,634.62  0.2616  5.2889  0.3796  0.3935  0.5699  0.6606  4.9380  ±2.84%     1318
   · Prepared with re-use  2,908.72  0.2303  5.0928  0.3438  0.3634  0.5307  0.5631  4.3680  ±2.58%     1455   fastest
   · Prepared each time    2,371.56  0.3070  6.8511  0.4217  0.4347  0.7030  0.9216  5.2908  ±3.33%     1186   slowest
```

[jraymakers]

Thanks! Interesting (but not surprising) that prepared statements have a slight overhead when not re-used.