docs: explain shared foundation migration

[imbajin]

Java plugin documentation still uses the 1.7.0 API. Add English and Chinese guidance for the planned 1.8.0 shared-foundation migration so developers can quickly find which imports, extension contracts and deployment steps affect them.

The guide starts with a reader-impact table and explains module ownership through a before/after illustration. A small IdGenerator example shows the import change; API/SPI tables cover custom implementations. A second illustration and a step-by-step upgrade section explain coordinated Server/PD/Store upgrades, matching metadata namespaces and the OLAP key change.

Compatibility details remain explicit: affected long-text indexes may need rebuilding, matching legacy OLAP rows remain readable, overwritten values cannot be recovered, and mixed-version rolling upgrades are unsupported. Plugin and contributor pages link the new guide. Dependency-inventory instructions prepare matching packaged libraries before collection. Historical-version routing does not advertise an unavailable migration page.

The implementation in apache/hugegraph#3270 has merged into Server master. This PR updates the bilingual migration status and contribution build prerequisites to match, including CI-friendly POM flattening for Maven 3.10. Matching service artifacts and downstream publication remain release prerequisites.

Validation: link validation and link-validator regression checks passed for the bilingual status and build-guide update. The site build, fresh artifact validation and visual checks run in PR CI.

[bitflicker64]

Blocking: yes. Summary: The guide's API and SPI claims match apache/hugegraph#3270, but its import table maps two whole packages to struct when some of their classes stay in core, so readers who apply those rows as written get imports that do not compile. Score 7/10. Evidence: static review of exact head 611003d against #3270 head eee539d49d67 (class locations via git ls-tree, HugeGraph.sameAs, GraphSerializer.writeIndex/readIndex, HugeElement.element()/wrapProperty, BaseVertex.TypeContext, pd.cluster default hg, Server cluster/usePD and graph pd.cluster options, regenerate_known_dependencies.sh and dependency_inventory.py); all latest-head checks pass.

[imbajin]

Blocking: yes. Summary: The migration guide omits the Server-wide PD namespace binding and the inventory command can retain stale jars on a reused checkout. Evidence: Paired implementation and inventory-collector locations are linked in the inline comments.

[imbajin]

Blocking: yes. Summary: The documented root Maven clean command can remove the PD and Store source trees through their distribution modules' broad Clean filesets. Evidence: the merged Server #3270 POMs point dist.dir at each parent module directory and configure it as a fileset; Maven Clean selects all paths when includes is empty, as linked in the inline comment.

[imbajin]

‼️ Critical: This root command invokes clean in the PD and Store distribution modules, whose POMs at merged Server #3270 set dist.dir to ${project.parent.basedir} and pass that entire module root as a Clean Plugin fileset without include or exclude patterns. Maven Clean's glob selector treats an empty include list as matching every path. This can delete the contents of hugegraph-pd/ and hugegraph-store/, including sources, POMs and uncommitted files, before packaging finishes. Please replace this with cleanup limited to generated distribution outputs, or narrow the paired Clean filesets before recommending the command; sync the Chinese guide. Evidence: PD dist.dir, PD Clean fileset, Store dist.dir, Store Clean fileset, Maven Clean selector.