sieve C9: record gen-c output determinism of t27c on 12 specs #6517

Boundary

• specs/compiler/theory/reproducible.t27 (new file; nothing else)

User Scenarios

• Sieve filter C9 (reproducible builds; reproducible-builds.org/docs/definition: the same source, environment and instructions give bit-identical artifacts) is UNKNOWN for t27c in specs/compiler/theory/sieve.t27 (epic EPIC: golden reference language -- the compiler sieve (TNF-style theorems), and t27 passes it #6488).
• The first piece of evidence anyone can collect with t27c alone is output determinism: does t27c gen-c give byte-identical output when run twice on the same spec?

Requirements

• Run t27c gen-c <spec> twice on each of these 12 specs and compare the two outputs with cmp:

  • specs/compiler/theory/sieve.t27
  • specs/compiler/theory/toolchain.t27
  • specs/compiler/theory/isa_round_trip.t27
  • specs/isa/ternary_encoding.t27
  • specs/fpga/assembler.t27
  • specs/tri/t27b/steward.t27
  • specs/queen/priority.t27
  • specs/compiler/core/t27core.t27
  • specs/math/constants.t27
  • specs/numeric/gf16.t27
  • specs/base/types.t27
  • specs/fpga/linker.t27

  Skip any spec that does not exist and record it as skipped.

• Record the result in the new module CompilerTheoryReproducible (specs/compiler/theory/reproducible.t27):

  • one pub fn identical(i: u32) -> bool and one pub fn spec_count() -> u32;
  • the master commit the run used, as a ; comment;
  • ; prose that explains this is output determinism of one t27c binary, not a bit-identical build of t27c itself. The latter needs two independent builds of t27c and stays open.

• Tests that call these functions at run time:

  • count the identical outputs and assert the number observed;
  • a negative control: two different specs whose gen-c outputs differ (cmp exits 1), recorded and asserted as different.

• Lines start with ; for prose (never a bare ;), ASCII only, English identifiers.

• The PR contains only .t27 and t27c-generated files (owner hard rule 2026-10-05).

• The PR says Closes #<this issue> and Refs #6488.

Success Criteria

• t27c typecheck specs/compiler/theory/reproducible.t27 prints Typecheck OK (0 errors.
• t27c gen-c specs/compiler/theory/reproducible.t27 > /tmp/r1.c && t27c gen-c specs/compiler/theory/reproducible.t27 > /tmp/r2.c && cmp /tmp/r1.c /tmp/r2.c exits 0.
• grep -c '^test ' specs/compiler/theory/reproducible.t27 prints 2 or more.
• git diff --name-only origin/master...HEAD lists only specs/compiler/theory/reproducible.t27.