refactor(consensus): remove namespace from client headers

[numinnex]

Removes the client-supplied namespace word from the VSR client request and reply headers. The server derives the routing target itself: the plane from operation, the partition group from the request body. At the admission boundary a request is promoted to the server-internal RoutedRequestHeader, which carries the resolved consensus group in the tail of the client header's reserved area (bytes 248..256).

This is a behavior change, not only a refactor. Name-addressed sends previously shipped headers pointing at partition (0, 0, 0), and the Rust default producer config could not send over VSR at all. With the namespace derived server-side, balanced and message-key partitioning now work over VSR for the Rust and Node SDKs.

Review follow-ups included here:

• RoutedRequestHeader::validate now enforces the same field rules as the client boundary (shared validate_request_fields), so a peer-wire Command2::Request frame cannot carry client = 0 into the client table's hard assert or replay a cached register reply via operation = Reserved.
• The client-to-routed promotion is an in-place retype (Message::into_routed) that zeroes only the eight group bytes instead of rebuilding the 256-byte header; the terminal validate stays.
• Cached client-table replies persist as raw ReplyHeader bytes, and this PR moves status from 224 to 216. Both persistence surfaces are versioned: the state-transfer artifact magic is bumped to ICT2, and MetadataSnapshot::decode refuses snapshots whose version differs from the current format version (bumped to 2).
• Layout pins: ReplyHeader::status at offset 216, RoutedRequestHeader::group at 248..256, and the data-bearing reserved prefix surviving promotion.

[codecov]

Codecov Report

❌ Patch coverage is 73.75796% with 206 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.52%. Comparing base (a3e256b) to head (b1fea62).

Files with missing lines	Patch %	Lines
core/shard/src/lib.rs	47.16%	54 Missing and 2 partials ⚠️
core/server-ng/src/dispatch.rs	37.73%	33 Missing ⚠️
core/simulator/src/client.rs	65.27%	25 Missing ⚠️
core/metadata/src/impls/metadata.rs	70.66%	22 Missing ⚠️
core/binary_protocol/src/consensus/header.rs	91.11%	12 Missing ⚠️
core/metadata/src/stm/snapshot.rs	77.27%	4 Missing and 1 partial ⚠️
core/partitions/src/state_transfer.rs	16.66%	5 Missing ⚠️
core/consensus/src/impls.rs	91.66%	4 Missing ⚠️
core/partitions/src/iggy_partitions.rs	75.00%	4 Missing ⚠️
core/shard/src/router.rs	75.00%	4 Missing ⚠️
... and 28 more

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3836      +/-   ##
============================================
- Coverage     76.95%   75.52%   -1.44%     
- Complexity     1021     1296     +275     
============================================
  Files          1366     1374       +8     
  Lines        176130   176464     +334     
  Branches     146173   146173              
============================================
- Hits         135547   133274    -2273     
- Misses        36707    39263    +2556     
- Partials       3876     3927      +51     

Components	Coverage Δ
Rust Core	75.92% <73.28%> (-0.15%)	⬇️
Java SDK	66.15% <ø> (+2.92%)	⬆️
C# SDK	47.00% <ø> (-29.01%)	⬇️
Python SDK	89.98% <ø> (ø)
PHP SDK	84.26% <ø> (ø)
Node SDK	96.25% <100.00%> (-0.08%)	⬇️
Go SDK	68.60% <100.00%> (-0.53%)	⬇️

Files with missing lines	Coverage Δ
core/binary_protocol/src/namespace.rs	90.90% <ø> (ø)
core/binary_protocol/src/version.rs	99.22% <ø> (ø)
core/consensus/src/client_table.rs	88.21% <ø> (ø)
core/consensus/src/lib.rs	0.00% <ø> (ø)
core/consensus/src/observability.rs	40.08% <100.00%> (ø)
core/consensus/src/plane_helpers.rs	96.24% <100.00%> (-0.03%)	⬇️
core/server-ng/src/bootstrap.rs	24.89% <ø> (ø)
core/server-ng/src/http/session.rs	85.80% <ø> (ø)
core/server-ng/src/http/wire.rs	95.93% <100.00%> (ø)
core/server-ng/src/partition_reconciler.rs	93.02% <100.00%> (ø)
... and 61 more

... and 129 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.

[hubcio]

• cached client-table replies persist as raw ReplyHeader bytes in the checkpoint and the state-transfer artifact, and this PR moved status 224 -> 216. correct today only by coincidence: every cached reply is metadata-plane so the old namespace word was 1<<63 with a zero low half, which the relocated status reads as 0. the state-transfer artifact has a magic tag but the msgpack snapshot reply bytes have no version marker at all. version both before this can bite.

• IggyNamespace::new masks (& STREAM_MASK) instead of rejecting and nothing caps live stream/topic count, so 4097 concurrently-live streams gives slab key 4096 which masks to stream 0 - a send to it lands in stream 0's partition and journals there. pre-existing (the resolve path is unchanged and string ids always masked server-side) so not a blocker here, but it's a silent cross-tenant write and this removes the last client-side fail-fast. worth its own issue: a fallible checked_new at the few sites that pack a runtime-derived key.

• worth noting in the description: this also enables balanced and message-key partitioning over VSR for the rust and node SDKs (the rust default producer config couldn't send over VSR before), so it's a behavior change, not only a refactor.

the core change is good - the deleted SDK packers re-derived a value the server computed anyway, and name-addressed sends were shipping headers pointed at partition (0,0,0). the gap is that the client boundary got hardened while the peer boundary stayed wire-trusting.

[numinnex]

/ready