feat: introduce container data structure for batch small file writing

[thuongle2210]

Description

Introduces Container — a new inode type that packs multiple small
files into a single shared block. N small files now require
1 inode + 1 block instead of N inodes + N blocks.

Resolve part of: Small files performance enhancement

Key Changes

1. New InodeContainer type with FileType::Container and
   InodeView::Container variant. Tracks per-file position via
   HashMap<String, SmallFileMeta>.
2. New end-to-end write flow: create_container → add_container_block
   → write → complete_container, on both master and worker.
3. Extend block-related methods (validate_add_block, get_block_locs,
   acquire_new_block) from InodeFile to InodeView to support both
   File and Container.
4. Generalize journal entries from File-specific (CreateFileEntry,
   CompleteFileEntry) to inode-generic (CreateInodeEntry,
   CompleteInodeEntry) with Container-aware replay.
5. Update RocksDB persistence: apply_new_block,
   apply_complete_inode_entry, get_file_locations, and tree
   reconstruction all handle Container.
6. Rename proto messages and RPC codes from *Batch* to *Container*.
7. Unit tests demonstrate batch writing consumes 1 block ID regardless
   of file count.

Performance Impact

• Metadata reduction: O(1) inodes and blocks per batch instead of O(N).
• Foundation for batch writing;

Future Work

1. Complete Container lifecycle (such as delete operation).
2. FsBatchWriter client interface.
3. Add advanced features: manual small file compaction or background compaction job integrated with streaming.
4. Move Container to middle/top folder levels (instead of leaf) if user demand is high.

[thuongle2210]

Dear @szbr9486 and @bigbigxu, could you please review this MR? Container packs large numbers of small files and dramatically reduces the number of inodes stored in metadata.

[szbr9486]

Dear @szbr9486 and @bigbigxu, could you please review this MR? Container packs large numbers of small files and dramatically reduces the number of inodes stored in metadata.尊敬的[姓名]，能否请您审核一下这个合并请求？容器会将大量小文件打包在一起，并大幅减少元数据中存储的 inode 数量。

Thank you for your excellent work. We have a holiday for Chinese new year in past few days, we will review this pr from today and wish you have a good day

[szbr9486]

Can this PR reduce the metadata memory usage in scenarios involving a large number of small files?

[lzjqsdd]

This PR can reduce the number of RPC calls and block allocations during writes. However, after the underlying data structure changes, it may not be very friendly for scenarios involving random reads of small files. If randomly read files are scattered across different blocks, we need to locate the offset first before reading. When the read cache hit rate is low, this could amplify I/O.

[thuongle2210]

Can this PR reduce the metadata memory usage in scenarios involving a large number of small files?

Maybe ContainerEntry (which is a similar file entry) can handle metadata memory consumption via lazy loading? This PR is quite large, so I think I'll add it in a subsequent PR.

[thuongle2210]

This PR can reduce the number of RPC calls and block allocations during writes. However, after the underlying data structure changes, it may not be very friendly for scenarios involving random reads of small files. If randomly read files are scattered across different blocks, we need to locate the offset first before reading. When the read cache hit rate is low, this could amplify I/O.

A container has exactly one block and SmallFileMeta stored offset.

[Copilot]

Pull request overview

This PR introduces a Container inode type to pack multiple small files into a single shared block, reducing metadata overhead from O(N) inodes+blocks to O(1) per batch. This addresses part of issue #433 regarding small file performance.

Changes:

• Adds new InodeContainer type and container lifecycle operations (create, add_block, complete)
• Generalizes journal entries from file-specific to inode-generic to support both File and Container types
• Renames proto messages and RPC codes from *Batch* to *Container* for clarity

Reviewed changes

Copilot reviewed 42 out of 45 changed files in this pull request and generated 17 comments.

Show a summary per file

File	Description
curvine-common/proto/common.proto	Adds Container file type and container_name field to FileStatusProto (breaking change)
curvine-common/proto/master.proto	Renames batch messages to container messages and adds container-specific responses
curvine-common/proto/worker.proto	Renames batch write messages to container write messages
curvine-server/src/master/meta/inode/inode_container.rs	New InodeContainer type with SmallFileMeta tracking
curvine-server/src/master/meta/inode/inode_view.rs	Extends InodeView enum with Container variant
curvine-server/src/master/meta/inode/inode_dir.rs	Adds container_index for mapping files to containers
curvine-server/src/master/meta/fs_dir.rs	Implements create_container and complete_container operations
curvine-server/src/master/meta/store/inode_store.rs	Updates RocksDB persistence for containers
curvine-server/src/master/journal/*	Generalizes journal entries for File and Container inodes
curvine-server/src/master/master_handler.rs	Implements create_container and complete_container RPC handlers
curvine-server/src/worker/handler/batch_write_handler.rs	Implements container block writing on worker side
curvine-client/src/file/fs_client.rs	Adds create_container and complete_container client APIs
curvine-client/src/block/*	Implements ContainerBlockWriter for batch small file writing
curvine-tests/tests/fs_test.rs	Adds test verifying 2 distinct block IDs for batch writes

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[thuongle2210]

Hi @lzjqsdd , I have just resolved all threads. Thanks!