fix(status,prompts): finish the #2432 residuals left by the merged #2580 family

[rysweet]

Context: the core #2432 work was superseded by the merged #2580 family

This branch began as a full parallel implementation of #2432 (eliminate silent
deterministic fallbacks + fix "zero active engineers"). While it was in flight,
main merged a comprehensive solution to the same problem:

• fix(ooda-brain): eliminate deterministic-default reasoner outcomes on parse-failure (#2580) #2588 — eliminated deterministic-default reasoner outcomes on parse-failure
  (explicit Err + brain_parse_error at the caller; parsers now consume a
  {"decision": …} / {"adjusted_urgency": …} envelope).
• fix(dashboard): report the TRUE live-engineer set so "Active Engineers" isn't stuck at zero (#2580) #2591 — reported the TRUE live-engineer set for the dashboard "Active
  Engineers" panel via a new operator_commands_dashboard::live_engineers module.

Those supersede this branch's parallel reasoner/dashboard implementation, so the
merge in this PR adopts main's version wholesale for all overlapping
production code and drops this branch's now-redundant zero_fallback_tests.rs
(main ships its own zero_fallback_2580_tests). No duplicate/competing
implementation is shipped.

What this PR now contains (the residual gaps #2580 left)

Main touched neither of these files, so both merge cleanly and are purely
additive:

1. status/provider.rs — finish design G4/G5 for the simard status surface

#2591 fixed the dashboard engineer count but left the status snapshot
(resources.live_engineers) on the buggy pgrep -f 'simard-engineer' (hyphen)
pattern — which never matches the real simard engineer run single-process …
(space) argv, so it undercounts. Read-only live-daemon evidence at reconcile time:

Signal	Observed
real …/simard engineer run single-process … subprocesses	present
live worktree dispatch claims (.simard-engineer-claim + live PID)	3 live
old status pattern pgrep 'simard-engineer' (hyphen)	never matches the space argv

resources.live_engineers now derives from the authoritative
count_live_engineer_claims (the single source of truth, design G4) — matching
the dashboard surface — and the fragile pgrep pattern is retired (design G5).
Covered by a new test (live_engineers_derives_from_live_worktree_claims).

2. The four .md embedded-fallback prompts — consistency with #2588's parsers

#2588 updated the recipe YAMLs to emit structured output but left the .md
embedded fallbacks (used when the on-disk prompt is absent; also the RustyClawd
path) still mandating the DECISION: marker / "Do NOT output JSON" — which
now contradicts #2588's parsers that consume a {"decision": …} envelope.
Refreshed ooda_decide.md, ooda_brain.md, ooda_orient.md, and
merge_readiness_judge.md to mandate the fenced JSON envelope with a required
decision field, preserving all pinned sentences (STATUS: ACHIEVED gate, six
lifecycle variants, DECISION marker, churn/stuck-loop).

3. parse_failure.rs — structured tracing

Converted a residual eprintln! (the gh-issue-filed success branch) to
tracing::info!, per the structured-tracing-only directive.

Validation

• Full cargo test --lib on the merged tree: 7031 passed, 0 failed (7 ignored
  are main's; the base_type_copilot integration tests skip when copilot is off
  PATH, as in CI).
• cargo fmt --check, pre-commit clippy --release -D warnings, and pre-push
  clippy --all-targets --all-features --locked -D warnings + the
  memory_consolidation race-subset all pass — no --no-verify, no --admin.
• git grep confirms no println!/eprintln! and no Bridge naming introduced.

Deploy

Does not touch the live daemon or ~/.simard (read-only inspection only). The
operator redeploys after merge.

Co-authored-by: Copilot 223556219+Copilot@users.noreply.github.com

[github-actions]

📊 Coverage Summary

Generated by cargo llvm-cov --workspace --summary-only (nightly, excluding test files)

Module	Lines	Covered	Coverage
Total	130505	108080	82.8%

_{Coverage data from CI run. Test files matching tests?/ are excluded from line counts.}