fix(shard): keep a parked prepare's stamp and op order on re-dispatch

[numinnex]

Partition prepares can arrive before their namespace is
materialised. Re-dispatch through the shard inbox lost each
frame's original epoch stamp and placed it behind newer frames.
A delete and recreate could therefore serve an old prepare
against the replacement, while the backup gap check could drop
a later operation before the parked prefix reached the plane.

Stage parked frames in a shard-local FIFO and carry their epoch
and park-pass provenance. A biased pump arm processes one
staged frame per iteration after ticks and before inbox work.
This preserves order without delaying heartbeats behind a
bounded queue, and loopback runs after every delivery so a solo
primary commits re-dispatched requests immediately. Shutdown
uses the same ordering.

Simulator materialisation now passes its resolved epoch and
wakes the pump after off-pump staging. Seeded three-replica and
solo-primary tests pin queue ordering, wakeup, loopback, and
deterministic replay. Quiescence rejects stranded redispatch
work.

The integration oracle now filters gap markers to the partition
plane, removes ambient RUST_LOG when an explicit logging level
is configured, and falls back to the server log when verbose
mode inherits stdout. Dropped or rejected staged frames are
counted, and recovery documentation reflects same-view repair.

The shared leader-redirection BDD fixture could elect replica 1
when a loaded runner delayed replica 0 during concurrent
startup. Start replica 1 only after replica 0 is healthy and use
a wider fixture heartbeat window so the scenarios test client
redirection instead of container scheduling.

[codecov]

Codecov Report

❌ Patch coverage is 93.73494% with 26 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.00%. Comparing base (328b289) to head (03d2590).
⚠️ Report is 3 commits behind head on master.

Files with missing lines	Patch %	Lines
core/shard/src/router.rs	71.42%	12 Missing and 4 partials ⚠️
core/shard/src/lib.rs	92.92%	7 Missing ⚠️
core/simulator/src/lib.rs	98.29%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #4005      +/-   ##
============================================
- Coverage     85.00%   85.00%   -0.01%     
  Complexity     1402     1402              
============================================
  Files          1225     1226       +1     
  Lines        180283   180538     +255     
  Branches     146587   146845     +258     
============================================
+ Hits         153248   153460     +212     
- Misses        22993    23016      +23     
- Partials       4042     4062      +20     

Components	Coverage Δ
Rust Core	85.89% <88.77%> (-0.01%)	⬇️
Java SDK	67.29% <ø> (ø)
C# SDK	75.38% <ø> (+0.01%)	⬆️
Python SDK	90.06% <ø> (ø)
PHP SDK	85.65% <ø> (ø)
Node SDK	96.22% <ø> (-0.03%)	⬇️
Go SDK	69.35% <ø> (+0.03%)	⬆️

Files with missing lines	Coverage Δ
core/server/src/partition_reconciler.rs	97.82% <100.00%> (-0.10%)	⬇️
core/shard/src/metrics.rs	86.55% <ø> (ø)
core/simulator/src/lib.rs	97.01% <98.29%> (+0.08%)	⬆️
core/shard/src/lib.rs	82.19% <92.92%> (-0.52%)	⬇️
core/shard/src/router.rs	78.40% <71.42%> (+0.13%)	⬆️

... and 37 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.