docs: refresh README.md and CLAUDE.md against current source

[marevol]

Summary

Brings README.md and CLAUDE.md back in line with the current source. The immediate trigger was the broken Maven Central badge, but the docs had drifted far enough that most of the README's code examples no longer compiled.

Badges

maven-badges.herokuapp.com now returns 404 (Heroku retired the free tier). Replaced with a shields.io Maven Central badge pointing at fess-crawler — the artifact users actually depend on — and added a javadoc.io badge. All badge and link targets verified to return 200.

Corrections

Every item below was checked against the source, not inferred:

Documented	Actual
15.2.0-SNAPSHOT	current release
JUnit 4	JUnit 5 (Jupiter) — there is no junit:junit dependency
crawler.crawlerContext / crawler.urlFilter	both protected; external code needs the delegating setters
container.singleton("httpClient", HcHttpClient.class, ...)	HcHttpClient is abstract — use Hc5HttpClient
setSocketTimeout	setSoTimeout
setMaxConnections	setMaxTotalConnections (init-param key is the singular maxTotalConnection)
setTrustAllCertificates	setIgnoreSslCertificate
crawlerContext.setDefaultIntervalTime(...)	does not exist — four delayMillis* knobs on DefaultIntervalController
FileTransformer.setMaxContentSize(...)	does not exist — limits live on ContentLengthHelper
accessResult.getContent()	getAccessResultData().getDataAsString()
OpenSearchDataService + setIndexName/setHostname/setPort	no no-arg constructor and no such setters; configured via OpenSearchCrawlerConfig
ExecutionTimeoutException under CrawlerSystemException	under ExtractException
MultipleCrawlingAccessException under CrawlerSystemException	under CrawlingAccessException
max line length 120	140 (per the shared formatter config in fess-parent)

The old "Basic Usage" example registered three components as singletons. StandardCrawlerContainer instantiates eagerly, so @Resource injection resolved to null and addUrl() NPE'd; a crawl needs roughly 25 components. The in-repo tests never caught this because they live in the same package as Crawler and so can reach its protected fields.

What replaced it

The Quick Start is now the Lasta DI bootstrap — SingletonLaContainerFactory.setConfigPath("crawler.xml") + init() — which gives a fully wired crawler in a few lines. It was compiled and run against the built jars: it crawls a seed page, follows child links, and reports results through DataService.

Every Java snippet in the README was extracted and compiled against fess-crawler and fess-crawler-lasta. The DI override examples were booted in a real container and their effects asserted.

Newly documented

• s3:// and gcs:// clients, and that S3Client falls back to the AWS default credentials provider chain when accessKey/secretKey are blank (IAM roles, IRSA, instance profiles)
• HtmlExtractor extracts default metadata by default; JSON-LD is opt-in
• Weight-based extractor selection and ExtractorBuilder for extracting from a stream without crawling
• Content-length caps are enforced during download, not after buffering
• fess-crawler-lasta declares jakarta.transaction-api as provided, so consumers must add it or startup fails with ClassNotFoundException: jakarta.transaction.Transactional$TxType
• ZipExtractor, TarExtractor, ApiExtractor and CommandExtractor ship in the jar but are not in the default DI
• Redefining extractorFactory in a DI file that includes crawler.xml replaces it wholesale and silently drops every default MIME mapping — verified: getExtractor("text/html") returns null. The README now shows appending instead, with a warning
• CI runs mvn -B package on a single JDK 21, so javadoc:jar and license:check gate the build even when mvn test is green
• JaCoCo reports coverage but enforces no threshold — the ">80%" line was aspirational
• Surefire runs test classes in parallel; JodExtractorTest is excluded; several tests need Docker

Notes

Documentation only — no source or build changes.