[ottonomy]

Closes #46

Stacked on #68 — review and merge that first (or merge the stack together).

Departures

None. Everything in "What to build" landed: ids stamped in the suite runner and
required on the type, the after-the-fact population removed, the summary's id
fallback removed, suite un-deprecated (check still deprecated),
SuiteSummary.verified renamed to hasFailures, the stability statement about
check ids and built-in phase tags written down, and the documentation sweep
(type comments, API guide, timing guide, README, CHANGELOG).

Two small extras inside the ticket's intent, both noted below: the API guide's
example gained the status.list-issuer row it was missing, and one line of
docs/architecture.md that spells out the CheckResult shape now lists id.

Decisions not in the issue

• computeId moved to src/compute-id.ts. The suite runner now needs it,
  and fold-results.ts already imports NOT_RUN_REASON_PREFIX from
  run-suites.ts, so importing the other way would have closed a cycle. The new
  module is where the "ids are API, fixed for 2.x" note lives.
  fold-results.ts re-exports computeId, so the barrel export and every
  existing import path are unchanged.
• The API guide's id list is in the order it was already in (core, proof,
  status, registry, recognition, parsing, then the three Open Badges bundles),
  not suite-execution order — smaller diff, same content.
• hasFailures is counts.failed > 0, i.e. any failure, fatal or not. The
  issue's wording ("true when any check in the suite failed") and the old
  field's inverse both read that way.
• The §1 example in the API guide gained a cryptographic.status.list-issuer
  row, because adding the missing id to the reference list while the worked
  example still showed the status suite with one check would have left the two
  disagreeing. Its summary row now reads "2 of 2 checks skipped".
• docs/architecture.md line 313 (CheckResult: { suite, check, outcome, … }) now includes id. The file is not in the ticket's documentation list,
  but that line states the shape of the type this change makes required; leaving
  it would have been a known-wrong line.
• Test fixtures for foldCheckResults need ids now. Rather than thread the
  phase through every call site, test/fold-results.spec.ts has a small
  FIXTURE_PHASES map from fixture suite id to phase, so mkResult stamps the
  same id the runner would — including custom, which is deliberately absent
  from the map so it produces unknown.custom.x.
• Four tests added to test/run-suites.spec.ts covering the acceptance
  item that the runner stamps every row: normal and not-run rows, an untagged
  suite (unknown. prefix), the synthetic .applies skip, and the .applies
  failure from a throwing predicate. The parse-failure id was already covered by
  an existing test in test/verify-credential.spec.ts.

Agent-Run: d76a817e-79b3-4e1c-bc5f-101e5d4c866e