GeezerVison

The build failures appear to be caused by #114751, so I think this is ready for review.

I believe the reason that the earlier version of this PR needed reverting was that it relied on the young gen being empty after a young collection. However if a finalizer creates new objects, then that assumption is false.

This version merges the young gen and increment before collection, so we don't care if the young gen isn't empty afterwards.

🤖 New build scheduled with the buildbot fleet by @markshannon for commit 912ebc1 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

!buildbot bigmem

🤖 New build scheduled with the buildbot fleet by @markshannon for commit 912ebc1 🤖

The command will test the builders whose names match following regular expression: bigmem

The builders matched are:

• AMD64 Windows11 Bigmem PR

🤖 New build scheduled with the buildbot fleet by @markshannon for commit 3c1ef42 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

!buildbot bigmem

🤖 New build scheduled with the buildbot fleet by @markshannon for commit 3c1ef42 🤖

The command will test the builders whose names match following regular expression: bigmem

The builders matched are:

• AMD64 Windows11 Bigmem PR

!buildbot bigmem

🤖 New build scheduled with the buildbot fleet by @markshannon for commit f4f04d6 🤖

The command will test the builders whose names match following regular expression: bigmem

The builders matched are:

• AMD64 Windows11 Bigmem PR

Note on the test failures:

• JIT / aarch64-unknown-linux-gnu/gcc (Debug) (pull_request) passes ~450 out of 460 tests then is timed out after an hour. I'm considering that a pass.
• Tests / CIFuzz (memory) (pull_request) is just broken and fails on everything at the moment.

This may have introduced a performance regression. See gh-117108.