Performance of LLVM 16 vs LLVM 15

commit

Filtered for unodb::db:

• micro_benchmark_key_prefix: 2% slowdown (unpredictable_prepend_key_prefix) to 0% slowdown (unpredictable_get_shared_length)
• micro_benchmark_n4: 15% slowdown (n4_full_scan/65535) to 3% speedup (n4_full_scan/100)
• micro_benchmark_n16: 9% slowdown (full_n16_tree_full_scan/64) to 19% speedup (full_n16_tree_sequential_delete/64)
• micro_benchmark_n48: 6% slowdown (full_n48_tree_random_delete/192) to 12% speedup (minimal_n48_tree_full_scan/8)
• micro_benchmark_n256:

Filtered for unodb::olc_db:

• micro_benchmark_key_prefix: 0% slowdown (unpredictable_get_shared_length) to 1% speedup (unpredictable_prepend_key_prefix)
• micro_benchmark_n4: 0% slowdown (shrink_node16_to_n4_sequentially/25) to 21% speedup (n4_full_scan/32768)
• micro_benchmark_n16: 20% slowdown (minimal_n16_tree_full_scan/16383) to 2% speedup (minimal_n16_tree_random_gets/16383)
• micro_benchmark_n48: 2% slowdown (minimal_n48_tree_full_scan/8) to 4% speedup (minimal_n48_tree_full_scan/4)
• micro_benchmark_n256: