refactor(tests): reduce complexity debt in test suites

[Trecek]

Summary

Refactor the test/support complexity debt tracked in #4993 across 37 files. Shared scope-selection, AST-scanning, fixture, fault-injection, and routing operations replace duplicated logic while preserving existing assertions, diagnostics, ordering, and cleanup.

• Lowered scores for 45 candidate identities, removed two unused helpers, and left one already-reduced candidate unchanged.
• Added the two planned regression tests and extended the existing hook-alias fixture.
• Kept production code, thresholds, and exemptions unchanged. The backend double's 33→1 caller score comes from moving its definition; its 32 methods remain at complexity 1.

The issue body contains all 48 candidate measurements and all introduced or relocated function measurements.

Local validation

• pre-commit run --all-files: passed.
• Affected gate: 6,341 passed, 104 skipped, 3 xfailed.
• Full configured local gate: 47,698 passed, 784 skipped, 34 xfailed; filtering disabled, configured Channel B exclusion retained.
• Supplemental checks against the captured base: 53 passed, covering three checks skipped in the full run for missing base context.
• Complexity checker: warning-only exit 0, with four extracted helpers above 10: source cascade (20), per-guard authority checks (17), shell outcome validation (11), and routing-target iteration (11).

Outstanding smoke validation

Both credentialed live smoke cases failed:

• Codex: unsuccessful session result before event normalization; the saved error dictionary does not establish the underlying refusal reason.
• Claude: one recognized subagent launch where the existing assertion requires at least two. Replaying its 130 raw events through the original and refactored normalizers produces identical 40-event output, including that one launch.

The smoke setup and live test body are unchanged. Assertions and launch-name sets were preserved, and no passing smoke result is claimed.

Refs #4993.