Add benchmark for the ilike operator, to compare it with like

[marregui]

Checklist

• [ ] CLA is signed

[marregui]

According to the benchmarks, 200 rounds, up to 3000 max rows:

• select "sourceIP" from uservisits where "sourceIP" like '%.%.%.%' limit 3000 Runtime in ms (mean: 47.494 ± 1.887), (min/max: 42.047 → 225.431) Percentile (50: 45.455 ± 13.613 (stdev)), (95: 53.412), (99.9: 225.431)
• select "sourceIP" from uservisits where "sourceIP" ILIKE '%.%.%.%' limit 3000 Runtime in ms (mean: 46.441 ± 1.861), (min/max: 40.880 → 204.743) Percentile (50: 44.376 ± 13.430 (stdev)), (95: 51.383), (99.9: 204.743)
• select "UserAgent" from uservisits where "UserAgent" like 'mozilla%' limit 3000 Runtime in ms (mean: 47.250 ± 0.485), (min/max: 42.748 → 69.374) Percentile (50: 46.156 ± 3.497 (stdev)), (95: 53.081), (99.9: 69.374)
• select "UserAgent" from uservisits where "UserAgent" ILIKE 'mozilla%' limit 3000 Runtime in ms (mean: 45.509 ± 0.306), (min/max: 41.318 → 57.075) Percentile (50: 45.126 ± 2.208 (stdev)), (95: 49.756), (99.9: 57.075)

I am happy to conclude that ILIKE performs slightly better than like.