Migrate to Apache Tika 4.0.0

[milamberspace]

Summary

Updates Apache Tika from 3.2.3 to the newly released 4.0.0 GA.

• Bump tika-core to 4.0.0 in bom-thirdparty.
• Drop the tika-parsers aggregator: it is packaging=pom and declared
  non-transitive, so it shipped no jar — it only dragged extra BOMs
  (grpc/azure/jetty/testcontainers) into dependency verification. JMeter
  bundles only tika-core; document parsing still relies on a
  user-supplied tika-app.jar at runtime.

API changes handled

• TikaConfig was removed in 4.x. In HTTPFileArg and
  ParseCurlCommandAction it was only used to set
  initializableProblemHandler="ignore", which is off the MIME-detection
  path (tika.detect uses detectors, not parsers). Replaced with a plain
  new Tika() and deleted the now-unused tika-config.xml.
• Parser.parse now requires a TikaInputStream. Document now wraps
  its byte array with TikaInputStream.get(...). Body text extraction is
  unchanged (BodyContentHandler still emits plain text by default).

Build / infrastructure

• Add trusted keys for the BOMs imported by tika-core's parent POM
  (com.azure, io.grpc, org.eclipse.jetty, org.testcontainers) in
  gradle/verification-metadata.xml. No jars from these are pulled — only
  their .pom signatures are validated.
• Update the expected release jar entry to tika-core-4.0.0.jar.
• Update the "Non-functional changes" changelog entry to Tika 4.x.

Testing

• ./gradlew :src:core:classes :src:protocol:http:classes style — pass.
• :src:protocol:http:test for TestHTTPFileArg, TestHTTPFileArgs,
  ParseCurlCommandActionTest — 30 passed, 0 failed.
• :src:dist:verifyReleaseDependencies — runtime classpath matches the
  expected jar list; signatures validated across the full runtime graph.

Note (out of scope)

Document.java still keys its "missing tika" hint off
java.class.path contains "tika-app". In 4.x tika-app ships as a zip
distribution rather than a fat jar, so that heuristic is now stale — left
untouched here as it concerns user runtime setup, not this build.

🤖 Generated with Claude Code