Easy way to respect BTreeMap's minimum node length

[ssomers]

Resolves#74834 the easy way.

Benchmarks (which are all biased/realistic, inserting keys in ascending order) say:

benchcmp r0 r3 --threshold 10
name r0 ns/iter r3 ns/iter diff ns/iter diff % speedup
btree::map::clone_fat_100_and_drain_all 156,137 178,180 22,043 14.12% x 0.88
btree::map::clone_fat_100_and_drain_half 121,160 142,412 21,252 17.54% x 0.85
btree::map::clone_fat_100_and_pop_all 176,110 196,380 20,270 11.51% x 0.90
btree::map::clone_fat_100_and_remove_all 215,075 237,690 22,615 10.51% x 0.90
btree::map::clone_fat_100_and_remove_half 142,377 164,035 21,658 15.21% x 0.87
btree::map::clone_fat_val_100_and_drain_half 60,372 70,638 10,266 17.00% x 0.85
btree::map::clone_fat_val_100_and_remove_half 64,124 75,858 11,734 18.30% x 0.85
btree::map::clone_slim_100_and_clear 2,183 2,838 655 30.00% x 0.77
btree::map::clone_slim_100_and_drain_all 3,652 4,350 698 19.11% x 0.84
btree::map::clone_slim_100_and_drain_half 3,320 4,443 1,123 33.83% x 0.75
btree::map::clone_slim_100_and_into_iter 2,154 2,836 682 31.66% x 0.76
btree::map::clone_slim_100_and_pop_all 3,372 4,051 679 20.14% x 0.83
btree::map::clone_slim_100_and_remove_all 5,111 5,820 709 13.87% x 0.88
btree::map::clone_slim_100_and_remove_half 3,259 4,198 939 28.81% x 0.78
btree::map::clone_slim_10k 244,160 276,820 32,660 13.38% x 0.88
btree::map::clone_slim_10k_and_clear 244,032 277,426 33,394 13.68% x 0.88
btree::map::clone_slim_10k_and_drain_all 384,035 425,290 41,255 10.74% x 0.90
btree::map::clone_slim_10k_and_drain_half 383,575 439,660 56,085 14.62% x 0.87
btree::map::clone_slim_10k_and_into_iter 241,980 274,260 32,280 13.34% x 0.88
btree::map::iter_0 1,733 1,487 -246 -14.20% x 1.17
btree::map::iter_100 2,714 3,707 993 36.59% x 0.73
btree::map::iter_10k 3,728 4,194 466 12.50% x 0.89
btree::map::iter_1m 5,570 6,581 1,011 18.15% x 0.85
btree::map::range_unbounded_unbounded 28,426 36,604 8,178 28.77% x 0.78
btree::map::range_unbounded_vs_iter 28,808 37,503 8,695 30.18% x 0.77
btree::set::clone_100_and_drain_half 2,935 3,458 523 17.82% x 0.85
btree::set::clone_100_and_remove_half 2,869 3,252 383 13.35% x 0.88
btree::set::clone_10k 205,255 232,898 27,643 13.47% x 0.88
btree::set::clone_10k_and_clear 204,650 231,412 26,762 13.08% x 0.88
btree::set::clone_10k_and_drain_all 319,003 354,305 35,302 11.07% x 0.90
btree::set::clone_10k_and_drain_half 332,200 374,400 42,200 12.70% x 0.89
btree::set::clone_10k_and_into_iter 204,652 229,792 25,140 12.28% x 0.89
btree::set::clone_10k_and_pop_all 318,360 351,400 33,040 10.38% x 0.91

r? @Mark-Simulacrum

[ssomers]

Not a work in progress, but I don't think the gloomy performance prediction makes it a genuine alternative to #75105.