refactor: extract shard allocator into a dedicated crate

[jayakasadev]

Which issue does this PR address?

Relates to #3315

Rationale

• decoupling core/server-ng from the legacy core/server crate

What changed?

• allocator moved to shard_allocator crate that only server and server-ng depend on

  • when I moved allocator to server_common, hwlocality leaked everywhere and broke the aarch64-musl build
    • simulator links with -nodefaultlibs
    • hwloc drags in glibc's libm.a, failing on an undefined reference to errno
  • server-ng had to import server crate to use shard allocator (server::shard_allocator) before this

• CpuAllocation/NumaConfig moved to a dedicated cpu_allocation leaf crate

  • depends only on serde, no hwloc
  • shard_allocator depends on cpu_allocation (not configs or server_common), so the allocator stays light and hwloc never leaks back into configs
  • configs::sharding re-exports them to keep the configs::sharding::* path stable for existing callers

• allocator lives in its own shard_allocator crate that only server and server-ng depend on

  • when I moved allocator to server_common, hwlocality leaked everywhere and broke the aarch64-musl build
    • simulator links with -nodefaultlibs
    • vendored hwloc drags in glibc's libm.a, failing on undefined references to glibc-internal math symbols
  • server-ng previously had to depend on the server crate just to reach the allocator (server::shard_allocator)

• CpuAllocation/NumaConfig moved to a dedicated cpu_allocation leaf crate

  • depends only on serde, no hwloc
  • shard_allocator depends on cpu_allocation (not configs or server_common), so the allocator stays light and hwloc never leaks back into configs
  • configs::sharding re-exports them to keep the configs::sharding::* path stable for existing callers

• shard_allocator carries a tiny, musl-gated build script that drops an empty libm.a stub on the link search path

  • vendored hwloc references cbrt, which forces the linker to resolve -lm; musl folds math into libc, so rust's musl sysroot ships no libm.a and -lm would otherwise fall through to glibc's libm.a (whose cbrt needs glibc-only __frexp/__ldexp, absent on musl)
  • the stub satisfies -lm harmlessly so cbrt resolves from musl's own libc, fixing the link independent of crate ordering

Local Execution

• Passed (macos); Linux/musl builds left to CI
• Pre-commit hooks ran

AI Usage

1. Tools: Cursor (Claude)
2. Scope: planned and performed the full relocation
3. Verification:

• ran fmt/sort/clippy + unit tests locally
• confirmed rg 'server::shard_allocator' core/server-ng/ is empty and that nothing was broken by the move

4. Yes, I can explain every line

[github-actions]

Thanks for the PR. It is labeled S-waiting-on-review and queued for review.

Slash commands (own line, regular comment) move it around the queue:

• /ready - back to S-waiting-on-review after addressing feedback
• /author - flip to S-waiting-on-author while you finish changes
• /request-review @user-or-team - request a reviewer

See CONTRIBUTING.md for details.

[codecov]

Codecov Report

❌ Patch coverage is 93.60731% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.15%. Comparing base (688d32a) to head (785f528).

Files with missing lines	Patch %	Lines
core/shard_allocator/src/lib.rs	8.33%	10 Missing and 1 partial ⚠️
core/cpu_allocation/src/lib.rs	98.55%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             master    #3578       +/-   ##
=============================================
- Coverage     74.35%   60.15%   -14.20%     
  Complexity      937      937               
=============================================
  Files          1257     1256        -1     
  Lines        130671   118917    -11754     
  Branches     106537    94829    -11708     
=============================================
- Hits          97156    71532    -25624     
- Misses        30434    44193    +13759     
- Partials       3081     3192      +111     

Components	Coverage Δ
Rust Core	57.18% <93.60%> (-17.88%)	⬇️
Java SDK	62.44% <ø> (ø)
C# SDK	71.40% <ø> (-0.71%)	⬇️
Python SDK	88.88% <ø> (ø)
PHP SDK	84.29% <ø> (ø)
Node SDK	91.26% <ø> (+0.14%)	⬆️
Go SDK	40.14% <ø> (ø)

Files with missing lines	Coverage Δ
core/configs/src/server_config/sharding.rs	100.00% <ø> (+14.55%)	⬆️
core/configs/src/server_config/validators.rs	78.39% <ø> (ø)
core/server-ng/src/bootstrap.rs	10.86% <ø> (ø)
core/server-ng/src/server_error.rs	92.10% <ø> (ø)
core/server/src/bootstrap.rs	82.23% <ø> (ø)
core/server/src/main.rs	63.25% <ø> (ø)
core/server/src/server_error.rs	0.00% <ø> (ø)
core/cpu_allocation/src/lib.rs	98.55% <98.55%> (ø)
core/shard_allocator/src/lib.rs	15.38% <8.33%> (ø)

... and 241 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[jayakasadev]

/request-review @hubcio