feat(cli): add clock-context diagnostic arm for issue 898

[wordingone]

Ember authority binding

goal_id: EMBER-02
workstream_id: EMBER-02A
next_executed_outcome: EMBER-02 first sufficiently pretrained clean-genesis 3B Ember

PR kind

kind:engineering

Linked issue or governing contract

Part of #898 and leaves #898 open.

Exact base SHA

4ffd491fcd5b944aa7f65be76bd57be4daf1db83

Exact reviewed head SHA

19d1a771a0605f4dfbf5007cb7508c82ae26f23b

Intended outcome

Add an opt-in diagnostic factor that pauses only the root ClockContext
one-second repaint so governed matched arms can test whether idle render cadence
drives the issue #898 memory slope.

Why this is one coherent PR

The exact environment gate and its mounted-App regression test form one bounded
diagnostic carrier. The test proves both the unchanged default cadence and the
flagged pause against the real reconciler rather than a substituted scheduler.

Affected areas

area:cockpit

Affected milestones

EMBER-02 — Three-billion-parameter foundation birth

This diagnostic affects the live milestone but grants no runtime-cure,
performance, milestone, or issue-closure credit.

Implementation summary

• Recognize only EMBER_DIAGNOSTIC_DISABLE_CLOCK_CONTEXT_TICK=1.
• Preserve the initial ClockContext value while suppressing only its root
  one-second interval.
• Leave the independent REPL liveness/heartbeat timer, resize fallback,
  fireball behavior, telemetry/activity pollers, and flag-absent behavior
  unchanged.
• Cover the default and flagged paths through a real mounted App observer.

Acceptance-clause mapping

1. Default compatibility. With the flag absent, the mounted root clock
   advances on its existing cadence.
2. Exact diagnostic factor. Only the exact value 1 suppresses the root
   clock interval and preserves its initial value.
3. Neighbor preservation. Focused neighboring suites cover resize,
   REPL liveness/fireball, and idle-spinner cadence.
4. Bounded claim. The carrier enables a later matched experiment; it does
   not itself identify or cure retained memory.

Local reproduction commands

From tools/ember-cli/src, run the new mounted-App test and the focused resize,
REPL liveness/fireball, home-screen fireball, and idle-spinner suites. Run the
project typecheck against both this head and the exact base, then run the normal
pre-commit repository guard.

Automated tests

• RED at base behavior: 1 passed, 1 failed; the exact flag test observed the
  expected second clock value.
• GREEN mounted-App suite: 2 passed, 0 failed, 5 expectations.
• Focused five-suite regression packet: 12 passed, 0 failed, 28 expectations.
• Project typecheck: 49 pre-existing errors at both base and head, byte-identical
  error set, with zero diagnostics in either changed path.
• Canonical pre-commit lifecycle audit and staged-scope repository guard: PASS.
• Exact staged whitespace check: PASS.

Executed evidence

No production diagnostic arm is claimed by this source PR. The P0/P1/P2
7,500-second matched-arm contract and adjudication thresholds are separately
preregistered under Leo authority; execution begins only from the unchanged
merged and installed binary.

Generated receipts and artifacts

No runtime receipt is checked into this source carrier. The later arms retain
their census, renderer, stdout/stderr, allocator, process-custody, and
self-hashed consumer evidence outside the repository under no-overwrite roots.

Known failures

The repository-wide typecheck currently reports the same 49 pre-existing
diagnostics at exact base and head. The carrier adds no changed-path diagnostic
and is diff-neutral against that baseline.

Unverified areas

• Public required checks and exact remote-head verification remain pending.
• No P0/P1/P2 arm has run from the merged installed binary.
• Whether idle render cadence or collection policy explains the retained-memory
  slope remains unadjudicated.

Claim boundary

This is an opt-in diagnostic carrier. Default production behavior is unchanged.
It grants no memory-cause finding, cure, benchmark, throughput, runtime PASS,
milestone, or issue-closure credit. Issue #898 remains open.

Review provenance

The two-path change was developed RED-first under Leo-approved bounded design.
Leo independently executed the RED, GREEN, focused neighboring regressions,
controlled base/head typecheck comparison, exact hash review, and canonical
hook chain. Fresh exact-head review and public required checks remain binding.

Rollback or revert procedure

Revert the eventual squash merge commit. This removes the diagnostic gate and
its test; the flag-absent default path is already identical to the predecessor.

Follow-up obligations that remain

• Verify pushed remote head, exact-head independent review, and required checks.
• Merge only after all publication gates are green.
• Run matched P0/P1/P2 arms sequentially on the same merged installed binary,
  with positive effective---smol evidence required before any P2 exclusion.
• Keep emberd absolute resource guard: OS-enforced commit/VRAM ceilings on every spawned run — kill the per-callsite preflight class #898 open pending a separately reviewed correction and governed evidence.