fix(eval): P5/P6 sieve-rank ladder replaces hardcoded FLAGGED_SIEVE_PREDS

- probe_report: dynamic lo/hi (no more hardcoded [101,200])
- main(): probe range derived from range_end ([range_end+1, range_end+1000])
- P6 = exact (no probe misses), P5(k) = errors match rank-k sieve signature
- _sieve_rank_signature: computes composites with all factors > p_k
- _compute_sieve_rank: identifies rank from model's error predictions
- build_report.py: probe_fig uses sieve rank for coloring
- Tests updated for P5/P6 codes (48/48 green)

1 files changed, 6 insertions, 5 deletions

diff --git a/tests/test_data.py b/tests/test_data.py
index 9f651fb..2fb131e 100644
--- a/tests/test_data.py
+++ b/tests/test_data.py
@@ -75,10 +75,11 @@ def test_make_batch_shapes_and_padding():
 
 
 def test_flagged_composites_are_composite():
-    # sanity: the diagnostic set really is composite, has no divisors <= 7, and covers
-    # the probe's candidate window [102, 211] (209 = 11*19 included — Addendum 3)
-    from src.eval import FLAGGED_SIEVE_PREDS
-    assert FLAGGED_SIEVE_PREDS == {121, 143, 169, 187, 209}
-    for n in FLAGGED_SIEVE_PREDS:
+    # sanity: the rank-4 sieve signature set is composite, has no divisors <= 7,
+    # and covers the probe's candidate window [102, 211] (Addendum 3/6)
+    from src.eval import _sieve_rank_signature
+    sig = _sieve_rank_signature(4, 101, 200)
+    assert sig == {121, 143, 169, 187, 209}, f"unexpected rank-4 signature: {sig}"
+    for n in sig:
         assert any(n % d == 0 for d in range(2, int(n ** 0.5) + 1))
         assert all(n % d != 0 for d in (2, 3, 5, 7))