Skip unchanged {{#each}} item subtrees during updates

[NullVoxPopuli-ai-agent]

Re-cut of #21512 (closed during the #21520 spike consolidation) onto current main, now with the full-suite verification that was outstanding: 9443 tests, 0 failures locally.

During revalidation, each {{#each}} item collects the tags its subtree consumed (via a tracking-frame finalizer on the item's try-frame); on later passes, an item whose collected tag validates clean is skipped entirely instead of walking every opcode of its subtree. A triviality gate (≤2 opcodes and no nested block) keeps the bookkeeping off items too small to profit. Measured standalone at ~1.9x on an 8x-throttled dbmon (walk-heavy, many clean items per frame); no observable behavior change — the same subtrees re-render for the same reasons, cheaper.

This was the campaign's first lever and its most reviewed artifact; it needs no RFC, no flag, and composes with (but does not require) the scheduler work in #21493/#21520.

🤖 Generated with Claude Code

[NullVoxPopuli-ai-agent]

Tracerbench PDF for a run with gc() before the first measurement and a measured gc() at the end: tracerbench-report.pdf

Ember's script time is 40 to 41% lower on update-every-10th, 21 to 22% on remove, and 29 to 31% on swap. Selecting the second row is 15.7% slower, as in both earlier runs. The final gc() takes as long as on main, so this PR leaves no extra garbage behind.

In the same setup, #21557 measured update -40%, remove -20 to -25%, swap -23%, selectSecondRow1 +9.9% slower, and a 6.6% slower first append.

phase	main script ms (8x)	script time	full phase
all phases	10199	-1.3%	no change
updateEvery10thItem1	87	-40.9%	-4.8%
updateEvery10thItem2	86	-40.3%	-3.8%
selectSecondRow1	127	+15.7% slower	+11.1% slower
removeFirstRow1	124	-21.9%	-4.9%
removeSecondRow1	122	-21.3%	-4.2%
swapRows1	107	-31.3%	-8.9%
swapRows2	107	-29.1%	-10.0%

How this was measured

• ember.js pnpm bench settings: tracerbench compare on smoke-tests/benchmark-app, 8x CPU throttle, headless, fidelity 50.
• Benchmark app change for this run, on control and experiment alike: gc() at the top of runBenchmark(), before the first mark, and a measured finalGc phase that calls gc() after clearItems4. Chrome runs with --js-flags=--expose-gc. Patch: benchmark-app-gc.patch
• Control is main 153364bb9a. Experiment is this PR on top of 153364bb9a (merged locally when the branch is behind).
• Chrome and tracerbench ran pinned to one CPU core. I also added --disable-background-networking,--disable-component-update to the Chrome flags and raised --sampleTimeout to 180 s.
• The PDF is tracerbench's own report. The only change is that its results folder reads results/<label> instead of a local path.
• Table values are the median of 50 per-round ratios. Tracerbench runs control and experiment back to back in each round, so both sides of a pair share the machine state. A value counts as a change when its bootstrap 95% CI excludes 0 and a Wilcoxon signed-rank test gives p < 0.05.
• "Script time" runs from the phase's Start mark to the end of the task that holds it: the click plus Ember's render. It leaves out layout, paint, the benchmark's row checks and idle time. Main-thread GC inside that window is removed.
• "Full phase" is tracerbench's own phase with main-thread GC removed.
• An earlier main-vs-main run (without the gc() calls) had false hits of up to +5% on script time.
• Raw results for every run: bench-reports branch.