The prototype's update_max_size function allocates a fake head node with __wt_calloc, performs the update, then frees it. This allocation/free happens on every insert/remove, adding CPU overhead and allocation churn. Eliminating this allocation improves performance.

Tasks

• src/block/block_ext.c — Rewrite update_max_size to use stack-allocated state instead of heap-allocated fake node (criterion: no malloc/calloc/free in update_max_size)
• src/block/block_ext.c — Verify that update_max_size logic is correct without the fake head: it should traverse from the changed offset upward through skiplist levels, recomputing max_size at each level (criterion: algorithm correctness preserved, validated by existing tests)
• test/suite/test_compact*.py — Run compaction test suite to ensure update_max_size works under real workload (criterion: python3 ../test/suite/run.py test_compact01 test_compact02 test_compact03 passes)

Files to create or modify

• src/block/block_ext.c — update_max_size function

Definition of Done

• update_max_size uses no heap allocations
• Algorithm correctness preserved (same results as prototype)
• Compaction tests pass, demonstrating real-world correctness