t27b vs t27c: the two runtime-assert counters disagree (string vs bool/int) #8074

Problem

t27c test-report (the reference) and t27b count runtime asserts differently, so the two backends disagree on which specs are vacuous:

• t27c counts a string comparison against a constant as a runtime assert, and does not count a bool or integer comparison.
• t27b counts the bool or integer comparison, and does not count the string comparison.

Example: specs/crons/999-multibots-telegraf-log-monitor.t27 has 4 string asserts and 1 bool assert. The reference reports 4 runtime asserts, t27b reports 1.

Effect: a spec can be vacuous in one backend and checked in the other (the specs in #8062 are vacuous only in t27b). The honest metric t27b_pass_where_reference_passes / reference_pass and the pass_vacuous ledger state both depend on this count.

Boundary

The decision of what counts as a runtime assert, written once in a .t27 spec, used by both counters.

Requirements

• One rule, stated in a .t27 spec with tests, decides whether an assert is a runtime assert (string, bool, integer, float).
• t27c and t27b report the same runtime-assert count on the same spec.

Success Criteria

• On specs/crons/999-multibots-telegraf-log-monitor.t27, t27c test-report and t27b print the same runtime-assert count.
• The lab run lists 0 specs that are vacuous in one backend and checked in the other.

Found by the vacuous-spec cleanup lane (#8052, #8054). Refs #6063

User Scenarios

• Given a spec with string comparisons against constants and bool/integer comparisons, when t27c and t27b both run test-report on that spec, then both backends report the identical runtime-assert count.

Requirements

• FR-001: A single .t27 spec MUST define the rule for what constitutes a runtime assert (string, bool, integer, float).
• FR-002: t27c MUST read the rule from that spec and count runtime asserts accordingly.
• FR-003: t27b MUST read the rule from that spec and count runtime asserts accordingly.
• FR-004: Both backends MUST produce the same runtime-assert count for any given spec.

Success Criteria

• t27c test-report specs/crons/999-multibots-telegraf-log-monitor.t27 and t27b specs/crons/999-multibots-telegraf-log-monitor.t27 print the same runtime-assert count (exit 0 for both, counts equal)
• Lab run lists 0 specs that are vacuous in one backend and checked in the other (0 specs divergent)