feat(server): dedup partition writes with per-group client table slices

[numinnex]

Implement deduplication for partition plane operations.

[numinnex]

Merge after #3958

[codecov]

Codecov Report

❌ Patch coverage is 90.50332% with 100 lines in your changes missing coverage. Please review.
✅ Project coverage is 73.01%. Comparing base (e77ee39) to head (949160c).

Files with missing lines	Patch %	Lines
core/shard/src/lib.rs	76.25%	35 Missing and 3 partials ⚠️
core/partitions/src/iggy_partitions.rs	37.93%	18 Missing ⚠️
core/partitions/src/iggy_partition.rs	90.47%	11 Missing and 1 partial ⚠️
core/partitions/src/state_transfer.rs	93.82%	5 Missing and 6 partials ⚠️
core/server/src/dispatch/partition.rs	80.76%	9 Missing and 1 partial ⚠️
core/consensus/src/client_table.rs	98.87%	2 Missing and 2 partials ⚠️
core/consensus/src/lib.rs	0.00%	3 Missing ⚠️
core/shard/src/router.rs	66.66%	1 Missing and 1 partial ⚠️
core/consensus/src/impls.rs	96.29%	0 Missing and 1 partial ⚠️
core/simulator/src/lib.rs	98.83%	1 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             master    #3959       +/-   ##
=============================================
- Coverage     85.10%   73.01%   -12.09%     
  Complexity     1402     1402               
=============================================
  Files          1236     1235        -1     
  Lines        181699   166022    -15677     
  Branches     148003   132330    -15673     
=============================================
- Hits         154629   121218    -33411     
- Misses        23047    40807    +17760     
+ Partials       4023     3997       -26     

Components	Coverage Δ
Rust Core	70.63% <90.50%> (-15.48%)	⬇️
Java SDK	67.29% <ø> (ø)
C# SDK	76.33% <ø> (ø)
Python SDK	90.06% <ø> (ø)
PHP SDK	85.65% <ø> (ø)
Node SDK	96.24% <ø> (+1.73%)	⬆️
Go SDK	69.35% <ø> (+0.06%)	⬆️

Files with missing lines	Coverage Δ
core/configs/src/server_config/defaults.rs	100.00% <100.00%> (ø)
core/configs/src/server_config/partition.rs	96.49% <100.00%> (+0.61%)	⬆️
core/server/src/boot/recovery.rs	95.96% <ø> (ø)
core/server/src/http/session.rs	95.54% <100.00%> (-0.14%)	⬇️
core/server/src/http/state.rs	88.42% <100.00%> (ø)
core/server/src/http/submit.rs	87.80% <100.00%> (-0.14%)	⬇️
core/server/src/partition_helpers.rs	72.56% <100.00%> (+0.07%)	⬆️
core/shard/src/metrics.rs	86.55% <ø> (ø)
core/consensus/src/impls.rs	91.50% <96.29%> (+0.06%)	⬆️
core/simulator/src/lib.rs	97.03% <98.83%> (+0.01%)	⬆️
... and 8 more

... and 262 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.

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs.

If you need a review, please ensure CI is green and the PR is rebased on the latest master. Don't hesitate to ping the maintainers - either @core on Discord or by mentioning them directly here on the PR.

Thank you for your contribution!

[numinnex]

/ready

[hubcio]

items on lines outside the diff:

• no shipped SDK preserves its client id across a reconnect and HTTP mints a fresh request id per call, so nothing reaches the dedup today - the PR body should say so and link the SDK follow-up.
• purge in iggy_partition.rs clears offsets, caches and consumer offsets but not the dedup slice, while converge_to_empty_after_failed_install clears it for the same reason. clear it in purge too.
• the decode_section byte-guard comment in state_transfer.rs is stale now that the dedup section follows: an over-stated group count fails as NonAscendingId or TrailingBytes instead of Truncated. fix the comment or subtract the trailing sections' minimum.

[numinnex]

/ready

[hubcio]

round 2 only fixed the doctest link, so most of round 1 still stands. two findings have no line in the diff to hang on:

ShardZeroCoordinator::seed_client_sequence in core/shard/src/coordinator.rs only ever raises the counter, and it seeds from live metadata ids at boot - a logged-out id is already gone from the table by then. so after a restart the same user can be handed the same client id while a peer primary still holds that id's slice watermark, and the first writes come back 201 without being written. minting ids that never repeat (boot nonce or view in the upper bits) closes it without touching the slice.

the PR body is still one line. worth saying which SDKs this reaches and what a client sees when a write is refused.

separate from this PR: the scripts/ci/storage-compat.sh fix in 85aa01e is ok.

[numinnex]

/ready