[lake] Propagate Hadoop configuration to Hudi writer

[Gezi-lzq]

Purpose

Closes #4585.

The Hudi write client currently uses HoodieFlinkEngineContext.DEFAULT, which does not inherit the Hadoop configuration assembled from the Hudi catalog settings. Consequently, S3A properties such as the endpoint and credentials are missing when Hudi accesses remote storage.

• No generative AI tools used
• Yes, OpenAI Codex was used to assist implementation, testing, and review

Brief change log

• Construct the Hudi write client's Flink engine context from the configured flinkConfig
• Preserve Hadoop filesystem properties, including S3A endpoint and credentials
• Add regression coverage for both COW/log and MOR/primary-key tables

Tests

• mise exec java@temurin-11 -- ./mvnw -pl fluss-lake/fluss-lake-hudi -am '-Dtest=org.apache.fluss.lake.hudi.**.*Test,org.apache.fluss.lake.hudi.**.*ITCase' -Dsurefire.failIfNoSpecifiedTests=false test
  • 113 tests passed, 0 failures, 0 errors, 0 skipped
• DevKit with Flink 1.20.3, Hudi 1.1.0, and RustFS S3A storage
  • Verified Hudi writes, appends, primary-key updates, native Hudi reads, and Fluss Union Reads
  • Followed the documented Hudi workflows for primary-key, log, partitioned, option passthrough, and automatic compaction scenarios

API and Format

No public API or storage format changes.

Documentation

No documentation changes. The existing $lake limitation for Hudi is unrelated and outside this PR's scope.

[luoyuxia]

cc @fhan688