[docs] Document Fluss S3 writing mechanism

[Copilot]

Purpose

Linked issue: N/A (research/documentation task)

This PR documents how Fluss writes data to S3, analyzing the architecture and data flow.

Brief change log

Research findings on Fluss S3 integration:

Architecture Layers:

• FileSystem → HadoopFileSystem → S3FileSystem → Hadoop S3AFileSystem → AWS S3

Key Components:

• S3FileSystemPlugin / S3AFileSystemPlugin: Handle s3:// and s3a:// URI schemes
• DefaultRemoteLogStorage: Orchestrates log segment copying to remote storage
• FsRemoteLogOutputStream: Buffered write stream to S3
• LogTieringTask: Background task for log tiering

Data Flow:

Local Log Segment → LogTieringTask → DefaultRemoteLogStorage → FsRemoteLogOutputStream → S3

Remote Directory Structure:

{remote.data.dir}/log/{db}/{table}_{tableId}/{bucketId}/{segment_uuid}/
    ├── {offset}.log
    ├── {offset}.index
    ├── {offset}.timeindex
    └── {offset}.writer_snapshot

Authentication: Supports static credentials (fs.s3a.access.key) or dynamic STS session credentials via S3DelegationTokenProvider.

Tests

N/A - research/documentation only

API and Format

No changes

Documentation

No code changes - this PR documents existing S3 integration for reference.

Original prompt

查看fluss中是怎么写入s3的

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.