refactor(dataset): Redirect multipart upload through File Service

[carloea2]

What changes were proposed in this PR?

• DB / schema

  • Add dataset_upload_session and dataset_upload_session_part tables to track multipart upload sessions, per-part status, and S3/LakeFS metadata.

• Backend (DatasetResource)

  • Partially new multipart upload API:

    • POST /dataset/multipart-upload?type=init → creates a LakeFS multipart session, stores it in DB, and returns an uploadToken.
    • POST /dataset/multipart-upload/part?token=...&partNumber=... → streams a single part to the presigned URL, with row-level locking and PENDING/UPLOADING/COMPLETED state transitions.
    • POST /dataset/multipart-upload?type=finish|abort → completes or aborts the LakeFS multipart upload and cleans up DB records.

  • Keep existing access control and dataset permissions enforced on all new endpoints.

• Frontend service (dataset.service.ts)

  • Main changes in multipartUpload(...):

    • Calls init to get uploadToken.
    • Uploads file parts via /multipart-upload/part streaming them with concurrency.

• Frontend component (dataset-detail.component.ts)

  • Use uploadToken to cancel/abort.

---

Any related issues, documentation, discussions?

Closes #4110

---

How was this PR tested?

• Manually uploaded large files via the dataset detail page (single and multiple), checked:

  • Progress, speed, and ETA updates.
  • Abort behavior (UI state + DB session cleanup).
  • Successful completion path (all parts COMPLETED, LakeFS object present, dataset version creation works).
  • Unit testing is missing

---

Was this PR authored or co-authored using generative AI tooling?

GPT partial use.

[carloea2]

@chenlica @aicam

According to our discussion I will look deeper to find a solution that does not involve creating new tables or mappings and rely on LakeFS/MinIO more.

[chenlica]

Thanks for checking the details. If you create a new PR, please mention it in this PR.