Let's land the ErrorRecovery work

[NullVoxPopuli]

Steps:

• baseline bench in js-framework-benchmark
  • needs to be updated to latest ember-source, as the VM has been upgraded
• link Error Recovery #1501 to ember-source (NOTE: additional followups here: Squashed feature handle error 2 #1528)
  pnpm link:all in glimmer-vm and pnpm link:glimmer-vm in ember-source
  • do the upgrade and PR it, as draft
    Spike error recovery integration emberjs/ember.js#20602
  • build ember-source and link to js-framework-benchmark
    Diff of released, vs, w/glimmer-vm#main so far: https://github.com/krausest/js-framework-benchmark/compare/master...NullVoxPopuli:glimmer-vm-main?expand=1
    Will add w/ the error-recovery PR
  • if the perf diff between current (not w/ glimmer-vm main) and post-error-recovery is "not significant" (in the wrong direction), we're good to move forward

Results:

Left to right:

• ember-source 5.7 (canary)
• ember-source 5.7 (canary + glimmer-vm's main branch)
• ember-source 5.7 (canary + glimmer-vm's error-recovery branch)

2323-12-12

ember-source is the same between both runs, (main)
only difference is glimmer-vm published vs main
still feels like margin of error though.
would need to compare with other folks' machines.

2023-12-11

NOTE: I need to use the same ember-source for both of these to reduce variables.
This bench had two variables:

• super old glimmer-vm vs glimmer-vm @ main
• ember-source 5.3 vs 5.6
• the current glimmer-vm on ember-source is much newer

ember-source 5.3 with glimmer-vm 0.84.2
vs
ember-source 5.6-alpha with glimmer-vm @ main