feat: Add server option "--allow-client-shm" (default false) controling client shared memory register/unregister

[yinggeh]

What does the PR do?

Introduces server option "--allow-client-shm=" (default false) to control client shm registration explicitly.
For example, if --allow-client-shm=false or unset (default)
✅ GET v2/systemsharedmemory[/region/${REGION_NAME}]/status
❌ POST v2/systemsharedmemory/region/${REGION_NAME}/register
❌ POST v2/systemsharedmemory[/region/${REGION_NAME}]/unregister
✅ GET v2/cudasharedmemory[/region/${REGION_NAME}]/status
❌ POST v2/cudasharedmemory/region/${REGION_NAME}/register
❌ POST v2/cudasharedmemory[/region/${REGION_NAME}]/unregister

if --allow-client-shm=true
✅ GET v2/systemsharedmemory[/region/${REGION_NAME}]/status
✅ POST v2/systemsharedmemory/region/${REGION_NAME}/register
✅ POST v2/systemsharedmemory[/region/${REGION_NAME}]/unregister
✅ GET v2/cudasharedmemory[/region/${REGION_NAME}]/status
✅ POST v2/cudasharedmemory/region/${REGION_NAME}/register
✅ POST v2/cudasharedmemory[/region/${REGION_NAME}]/unregister

Checklist

• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Populated github labels field
• Related issues are referenced.
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• feat

Related PRs:

Where should the reviewer start?

Test plan:

• CI Pipeline ID:
  48256373

Caveats:

Background

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: #xxx

[whoisj]

Looks good to me. Just have 1 ask before approval.

[whoisj]

Look good. Only ask is to correct the markdown to use the more modern markdown for alerts.

Also, a thought about using a constant in place of repeated literals.

[Copilot]

Pull request overview

Adds a new server CLI switch (--allow-client-shm=<bool>, default false) to explicitly control whether clients may register/unregister shared-memory regions, while keeping shared-memory status queries available and leaving backend-internal shared memory unaffected.

Changes:

• Adds --allow-client-shm parsing/plumbing and passes the value into SharedMemoryManager.
• Rejects HTTP + gRPC shared-memory register/unregister calls when client SHM is disabled (default).
• Updates docs and QA to cover the new default behavior and to enable SHM explicitly for SHM-dependent tests.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/shared_memory_manager.h	Adds constructor parameter + accessor for client-SHM enablement.
src/main.cc	Instantiates SharedMemoryManager with the parsed CLI flag.
src/http_server.cc	Blocks SHM register/unregister endpoints when client SHM is disabled.
src/grpc/grpc_server.cc	Blocks SHM register/unregister RPCs when client SHM is disabled.
src/common.h	Introduces a shared error string for the disabled-client-SHM case.
src/command_line_parser.h	Adds allow_client_shm_ to server parameters + option grouping storage.
src/command_line_parser.cc	Adds CLI option definition/group and parses --allow-client-shm.
qa/L0_shared_memory/test.sh	Adds coverage for default-disabled behavior; enables SHM for existing tests.
qa/L0_shared_memory/shared_memory_test.py	Adds test asserting register/unregister are rejected by default.
qa/L0_cuda_shared_memory/test.sh	Adds coverage for default-disabled behavior; enables SHM for existing tests.
qa/L0_cuda_shared_memory/cuda_shared_memory_test.py	Adds test asserting CUDA SHM register/unregister are rejected by default.
qa/L0_sequence_batcher/test.sh	Enables --allow-client-shm=true when SHM-related subtests run.
qa/L0_infer/test.sh	Enables --allow-client-shm=true when SHM-related subtests run.
qa/L0_batcher/test.sh	Enables --allow-client-shm=true for CUDA-SHM tests; adjusts warmup + fixes typo.
docs/protocol/extension_shared_memory.md	Documents the new default-disabled requirement for client SHM usage.

---

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

[whoisj]

LGTM