Massive performance regression since commit fbd332d

Report

• I have searched existing issues and this is not a duplicate

The recent move to "OOP" patterns and code structures has heavily impacted performance as can be seen when running benchmarks:

TAG [1.8.0]
Stack with flex: median: 0.027000 ms, stddev: 0.005840 ms
Align stretch in undefined axis: median: 0.027000 ms, stddev: 0.005836 ms
Nested flex: median: 0.436000 ms, stddev: 0.067741 ms
Huge nested layout: median: 36.938000 ms, stddev: 2.847124 ms

COMMIT [fbd332d]
Stack with flex: median: 0.042000 ms, stddev: 0.007671 ms
Align stretch in undefined axis: median: 0.036000 ms, stddev: 0.009329 ms
Nested flex: median: 0.748000 ms, stddev: 0.108319 ms
Huge nested layout: median: 50.196000 ms, stddev: 3.076368 ms

TAG [1.7.0]
Stack with flex: median: 0.010000 ms, stddev: 0.006567 ms
Align stretch in undefined axis: median: 0.012000 ms, stddev: 0.002526 ms
Nested flex: median: 0.152000 ms, stddev: 0.038512 ms
Huge nested layout: median: 12.710000 ms, stddev: 1.497397 ms

As far as I understand mostly due to unnecessary copies all over the place.
I'm not sure if this is a known issue, but I suggest taking a look at it before it becomes too hard to optimize/refactor.

Start by running the mentioned benchmarks and comparing TAG 1.7.0, TAG 1.8.0, and commit fbd332d. Trace the recent OOP-related changes and investigate the reported unnecessary copies; done means the regression is explained and benchmark results show the performance issue has been addressed.