Consolidate repeated crate-local test setup fixtures

[forkwright]

Finding

Akroasis's test surface repeatedly reconstructs the same setup values inside files that were split primarily to satisfy file-length limits. These are not independent protocol vectors or hardcoded expected values whose duplication catches production drift; they are fixture constructors expected to behave identically.

A normalized exact-function-body scan on the verified revision found repeated setup families across Kerykeion, Semaino, Kryphos, Tekmerion, Syntonia, and the Akroasis caller/CLI tests.

Verified against main c8e671845adca3e5b57a205d08aca31eadbc6000.

Evidence

Representative exact copies include:

• crates/kerykeion/src/collector_tests.rs:8-15 and collector_tests_attribution.rs:8-15: identical make_config.
• crates/kerykeion/src/processor_tests.rs:8-13 and processor_tests_attribution.rs:11-16: identical make_processor. The files are explicitly split because the primary test files reached the 800-line threshold.
• crates/semaino/src/aggregator.rs:279-291 and pipeline.rs:350-362: identical rf_signal builders.
• crates/kryphos/tests/tamper_log_signing.rs:202-211 and crates/kryphos/src/storage_tests.rs:761-770: identical code for removing installation-identity fields from a vault header.
• crates/tekmerion/src/tamper_log_tests.rs:9-11,32-38 and tamper_log_codec_tests.rs:7-18: identical test_key and config_kind fixtures.
• crates/syntonia/tests/workflow_concurrency_fitness.rs:34-46 and gate_feature_coverage.rs:21-33: identical workspace-root and file-read helpers.
• crates/akroasis/tests/caller_contract.rs:270-290, caller_receipt_wire.rs:124-134, and sibling caller tests repeat fixed timestamp and policy/schema epoch constructors.
• crates/akroasis/src/mesh/mod.rs, radio/mod.rs, and vault/mod.rs each define the same Clap test parser wrapper over TestCli::parse_from.

The exact-body scan also found intentionally independent expected values and trivial trait mocks; those are excluded from this finding.

Why this matters

The purpose of splitting a test file is readability and lint compliance, not creating a second owner for setup policy. As fixtures evolve, one split file can run with a different local-node identity, baseline signal shape, legacy-header mutation, workspace-root assumption, or caller epoch than its sibling.

The duplication also obscures the actual assertion. A test reader must verify repeated setup before reaching the behavior under test, and a correction to setup requires a repository-wide search rather than one fixture edit.

Desired correction

Create narrow #[cfg(test)] support modules inside the owning crate or parent module:

• Kerykeion collector/processor fixtures shared by their split test modules;
• Semaino signal builders;
• Kryphos legacy-header mutation helpers;
• Tekmerion tamper-log fixtures;
• Syntonia fitness-test filesystem helpers; and
• Akroasis caller/CLI test constructors.

Do not create a fleet-wide generic test framework. Keep cryptographic vectors, wire-format bytes, independently calculated expected values, and adversarial fixtures local where their independence is load-bearing.

Done when:

• each listed setup family has one implementation in its owning crate;
• split test files import fixtures rather than copy them;
• fixture defaults and knobs are explicit;
• expected-value vectors intended to detect production drift remain independent; and
• an exact-body scan no longer reports these setup constructors as duplicate families.