fix: Scrub class for better detection of global data frames

[krlmlr]

Code generation for TPC-H currently doesn't work. This is an attempt at fixing this, but incomplete.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 1f0ba84 is merged into main:

• ✔️001_tpch_01: 21.7ms -> 22ms [-3.25%, +5.6%]
• ✔️001_tpch_02: 64.8ms -> 64.7ms [-1.36%, +1.03%]
• ✔️001_tpch_03: 40.5ms -> 40.9ms [-1.07%, +2.96%]
• ✔️001_tpch_04: 22.2ms -> 22.4ms [-1.36%, +2.97%]
• ✔️001_tpch_05: 58.1ms -> 58.3ms [-0.42%, +1.06%]
• ✔️001_tpch_06: 11.6ms -> 11.7ms [-1.82%, +3.47%]
• ✔️001_tpch_07: 71.7ms -> 71.9ms [-0.57%, +1.22%]
• ✔️001_tpch_08: 98.8ms -> 98.9ms [-1.37%, +1.55%]
• ✔️001_tpch_09: 69.2ms -> 69.7ms [-0.79%, +2.04%]
• ❗🐌001_tpch_10: 42.5ms -> 42.9ms [+0.03%, +1.55%]
• ✔️001_tpch_11: 38.3ms -> 38.3ms [-1.62%, +1.35%]
• ✔️001_tpch_12: 21ms -> 20.6ms [-3.85%, +0.55%]
• ✔️001_tpch_13: 18.3ms -> 18.3ms [-3.43%, +2.73%]
• ✔️001_tpch_14: 16.6ms -> 16.8ms [-0.4%, +3.51%]
• ✔️001_tpch_15: 36.8ms -> 36.4ms [-2.95%, +0.98%]
• ✔️001_tpch_16: 37.5ms -> 37.6ms [-2.84%, +3.22%]
• ✔️001_tpch_17: 33.7ms -> 33.4ms [-3.02%, +1.4%]
• ✔️001_tpch_18: 23.5ms -> 23.4ms [-3.11%, +2.27%]
• ✔️001_tpch_19: 46.4ms -> 46.5ms [-1.34%, +1.72%]
• ✔️001_tpch_20: 52ms -> 52.3ms [-0.65%, +1.92%]
• ✔️001_tpch_21: 73.3ms -> 73.4ms [-1.31%, +1.73%]
• ✔️001_tpch_22: 47.9ms -> 48.2ms [-1.21%, +2.15%]
• ✔️010_tpch_01: 78ms -> 75.7ms [-6.64%, +0.65%]
• ✔️010_tpch_02: 69.2ms -> 70.1ms [-0.42%, +3.27%]
• ✔️010_tpch_03: 58.7ms -> 57.3ms [-8.52%, +3.78%]
• ✔️010_tpch_04: 41.5ms -> 41.7ms [-1.11%, +2.18%]
• ✔️010_tpch_05: 87.9ms -> 87ms [-2.02%, +0.08%]
• ✔️010_tpch_06: 29.5ms -> 29.9ms [-8.4%, +11.12%]
• ✔️010_tpch_07: 100ms -> 101ms [-0.68%, +1.62%]
• ✔️010_tpch_08: 126ms -> 126ms [-0.55%, +1.69%]
• ✔️010_tpch_09: 115ms -> 114ms [-3.1%, +1.36%]
• ✔️010_tpch_10: 72.2ms -> 75.5ms [-1.43%, +10.79%]
• ✔️010_tpch_11: 37.4ms -> 37ms [-4.31%, +2.47%]
• ✔️010_tpch_12: 52.2ms -> 50.8ms [-7.5%, +2.12%]
• ✔️010_tpch_13: 50.7ms -> 50.6ms [-3.42%, +3.04%]
• ✔️010_tpch_14: 33.7ms -> 33.4ms [-3.56%, +1.84%]
• ✔️010_tpch_15: 53.4ms -> 53.5ms [-4.31%, +4.51%]
• ✔️010_tpch_16: 38.6ms -> 38.4ms [-3.32%, +2.18%]
• ✔️010_tpch_17: 52.9ms -> 51.3ms [-7.77%, +1.67%]
• ✔️010_tpch_18: 55.2ms -> 53.3ms [-11.18%, +4.11%]
• ✔️010_tpch_19: 82.5ms -> 84ms [-2.4%, +6.16%]
• ✔️010_tpch_20: 65ms -> 62.7ms [-7.39%, +0.24%]
• ✔️010_tpch_21: 235ms -> 235ms [-3.69%, +3.03%]
• ✔️010_tpch_22: 53.4ms -> 52ms [-5.18%, +0.04%]
• ✔️100_tpch_01: 317ms -> 323ms [-16.51%, +19.99%]
• ✔️100_tpch_02: 125ms -> 129ms [-7.33%, +13.91%]
• ✔️100_tpch_03: 171ms -> 169ms [-5.24%, +3.13%]
• ✔️100_tpch_04: 150ms -> 149ms [-4.6%, +3.1%]
• ✔️100_tpch_05: 262ms -> 252ms [-12.9%, +5.33%]
• ❗🐌100_tpch_06: 92.6ms -> 103ms [+9.85%, +13.04%]
• ✔️100_tpch_07: 224ms -> 230ms [-5.45%, +11.27%]
• ✔️100_tpch_08: 253ms -> 256ms [-7.42%, +9.21%]
• ✔️100_tpch_09: 327ms -> 341ms [-14.31%, +23.09%]
• ✔️100_tpch_10: 204ms -> 209ms [-5.31%, +10.15%]
• ✔️100_tpch_11: 81.9ms -> 81ms [-2.84%, +0.71%]
• ✔️100_tpch_12: 179ms -> 178ms [-11.51%, +9.73%]
• ✔️100_tpch_13: 309ms -> 302ms [-9.29%, +4.9%]
• ✔️100_tpch_14: 112ms -> 110ms [-11.69%, +7.52%]
• ✔️100_tpch_15: 200ms -> 199ms [-11.12%, +9.32%]
• ✔️100_tpch_16: 121ms -> 118ms [-6.52%, +1.81%]
• ❗🐌100_tpch_17: 178ms -> 188ms [+0.74%, +10.18%]
• ✔️100_tpch_18: 189ms -> 192ms [-8.31%, +10.71%]
• ✔️100_tpch_19: 279ms -> 273ms [-16.7%, +12.56%]
• ✔️100_tpch_20: 178ms -> 172ms [-8.35%, +1.41%]
• ✔️100_tpch_21: 1.29s -> 1.31s [-6.13%, +9.12%]
• ✔️100_tpch_22: 150ms -> 148ms [-7%, +4.11%]

Further explanation regarding interpretation and methodology can be found in the documentation.