Add PathPatternNode wildcard cache benchmark

[Caideyipi]

Description

Adds an opt-in microbenchmark for the compiled non-trivial wildcard cache introduced by #18672.

• Reproduces the pre-change compile-on-every-match path.
• Compares it with PathPatternNode.getMatchChildren, alternating execution order by round.
• Checks that both paths return the same number of matches.
• Reports median latency, throughput, speedup, and latency reduction for configurable wildcard counts.
• Stays skipped in normal UT runs unless -Diotdb.path.pattern.node.perf.enabled=true is set.

Example:

mvn -pl iotdb-core/node-commons -Dtest=PathPatternNodePerformanceTest -Diotdb.path.pattern.node.perf.enabled=true test

The workload can be tuned with the iotdb.path.pattern.node.perf.* system properties documented by the test skip message.

Local benchmark results

Run on 2026-09-21 with Windows 11 (build 26200), JDK 17.0.15, and an Intel Core i9-13900H. The command used for both revisions was:

mvn -pl iotdb-core/node-commons -Dtest=PathPatternNodePerformanceTest -Diotdb.path.pattern.node.perf.enabled=true -Diotdb.path.pattern.node.perf.wildcard-counts=1,10,100 -Diotdb.path.pattern.node.perf.query-count=1000 -Diotdb.path.pattern.node.perf.warmup-lookups=2000 -Diotdb.path.pattern.node.perf.lookups=20000 -Diotdb.path.pattern.node.perf.rounds=7 test

The PR base is 9be80ead1dc (the #18672 cache implementation); the PR head is e78adaf878e. The benchmark's legacy column emulates the pre-#18672 implementation, while the cached column calls PathPatternNode.getMatchChildren:

wildcard children	pre-#18672 path (ns/lookup)	cached path (ns/lookup)	cached throughput (lookups/s)	speedup	latency reduction
1	1,217.7	205.0	4,879,239	5.94x	83.17%
10	1,767.5	348.5	2,869,193	5.07x	80.28%
100	16,148.7	2,800.8	357,039	5.77x	82.66%

For the cached path alone, the same run on the PR base versus the PR head was:

wildcard children	base 9be80ead1dc (ns/lookup)	head e78adaf878e (ns/lookup)	head vs. base
1	198.7	205.0	+3.2%
10	291.9	348.5	+19.4%
100	2,684.4	2,800.8	+4.3%

#18685 only adds the benchmark test (the two-line OPC UA reference cleanup in the head is unrelated to PathPatternNode); the base/head differences above are single-run microbenchmark noise, not a production-path change. The signal is the comparison against the emulated pre-#18672 path: about 5-6x higher throughput and 80-83% lower lookup latency with the compiled-pattern cache.

Testing

• mvn spotless:apply -pl iotdb-core/node-commons - passed.
• Targeted PathPatternNodePerformanceTest with the command above - passed on the PR head (1 test, 0 failures).
• The same benchmark passed on a temporary worktree at the PR base; the benchmark source and the head's two-line OPC UA reference cleanup were applied there only to bypass an unrelated compile issue.
• Checkstyle, Spotless, compilation, and Surefire phases in the targeted Maven run - passed.

---

This PR has:

• been self-reviewed.
• added comments explaining the benchmark intent.
• kept the performance test disabled during regular unit-test runs.

---

Key changed/added classes

• PathPatternNodePerformanceTest