refactor(crawler): tighten CrawlerThread hot path

[marevol]

Summary

Reduce stream/lambda allocations on per-crawl hot paths and small init code, mirroring the same direction as fess#3134. No behavior changes.

Changes

• CrawlerThread.storeChildUrls: Convert the stream().filter(...).map(...).collect(Collectors.toList()) pipeline into a single for loop. Pre-size both the dedup HashSet and the resulting ArrayList from childUrlList.size(). Reuses the URL value via a local variable instead of three repeated d.getUrl() calls. This runs once per crawled page that yields children.
• CrawlerClientCreator.register(...): Replace clientMap.entrySet().stream().forEach(...) and clientFactoryList.forEach(...) with enhanced for loops. Both methods are synchronized and run during client registration; the change just removes lambda/stream object allocations.
• UrlFilterImpl.init: Replace cachedXxxSet.stream().collect(Collectors.toList()) with new ArrayList<>(set). Drops the java.util.stream.Collectors import.

What was intentionally NOT changed

• The set/list types (LinkedHashSet, HashSet, ArrayList) are kept the same; only the way they are populated changes.
• No public/protected method signatures change.
• The redirect/HEAD logic in CrawlerThread is untouched.

Test plan

• mvn compile — succeeds
• mvn test -Dtest=UrlFilterImplTest,CrawlerThreadTest — 34/34 pass (no CrawlerClientCreatorTest exists in this repo)
• mvn formatter:format && mvn license:format — applied, clean
• Manual crawl smoke test against a small site

🤖 Generated with Claude Code