[Feature] client support bypass server for HStore writing

### Search before asking

- [X] I had searched in the [feature](https://github.com/apache/incubator-hugegraph-toolchain/issues?q=is%3Aissue+label%3A%22Feature%22) and found no similar feature requirement.

### Feature Description (功能描述)

first huge-client support bypass server for HStore writing

# Tasks

- [ ] loader write vertices/edge direct to HStore by spark loader
- [ ] loader write vertices/edge generate SST file

### Are you willing to submit a PR?

- [X] Yes I am willing to submit a PR!

### Code of Conduct

- [X] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

Start by locating the huge-client and Spark loader entry points, then trace how vertices and edges are currently written through the server. The issue’s done criteria are direct HStore writes from the loader and generation of SST files, with both tasks covered and verified by the project’s existing checks.

Written by the indexing model from the issue text.