[llvm] [ADT] ImutAVLFactory: avoid overhead of vector for freelist (PR #196534)

[Christoph Cullmann via llvm-commits]

christoph-cullmann wrote:

Tried with std::vector like before, can not see a re-producable speed diff.
(clang-tidy runs seem for me not totally stable runtime wise, with and without any of these patches, the memory usage changes are stable)

Memory usage (as expected) the same as with the SmallVector.
I would consider the current state of this patch ok from my side.

Feedback welcome.

https://github.com/llvm/llvm-project/pull/196534