Exclusive/complementary constraint resolution for group children

[ruggi]

Fixes #4281

Problem:

It's possible to over-specify group child constraints, e.g. setting both left, right, and width.

Fix:

1. When setting a dimension constraint (w/h), if both the side complements (tlbr) are set, remove the oldest constraint
2. When setting a side constraint (tlbr), if both the dimension complements (w/h) are set, remove the dimension complement (so side pins win)

[github-actions]

Try me

[relativeci]

Job #8339: Bundle Size — 62.7MiB (~+0.01%).

6f34c44(current) vs 5957b21 master#8338(baseline)

Warning

Bundle contains 64 duplicate packages - View duplicate packages

Bundle metrics (1 change)

	Current Job #8339	Baseline Job #8338
Initial JS	34.97MiB(~+0.01%)	34.97MiB
Initial CSS	0B	0B
Cache Invalidation	18.38%	18.38%
Chunks	27	27
Assets	31	31
Modules	3970	3970
Duplicate Modules	422	422
Duplicate Code	30.83%	30.83%
Packages	409	409
Duplicate Packages	64	64

Bundle size by type (1 change)

	Current Job #8339	Baseline Job #8338
CSS	0B	0B
Fonts	0B	0B
HTML	11.43KiB	11.43KiB
IMG	0B	0B
JS	62.69MiB (~+0.01%)	62.69MiB
Media	0B	0B
Other	0B	0B

---

View job #8339 report View feat/exclusive-constraint-props branch activity

[bkrmendy]

since we set the data-constraints prop via Utopia, I think it'd be worth it to validate its value getSafeGroupChildConstraintsArray so that we only take into account the valid values, and ignore whatever else that might have made its way into the prop (and get rid the anys stemming from getSafeGroupChildConstraintsArray

[ruggi]

makes sense, done in 6f34c44 (#4282)

[github-actions]

Performance test results: