Fix D9: z-score thresholds from two-tailed to one-tailed

Change Z_90 from 1.645 to 1.2816 and Z_95 from 1.96 to 1.6449 in
repness.py, matching Clojure's one-tailed z-scores (stats.clj) and
resolving an internal inconsistency with Python's own stats.py which
already used the correct values.

The proportion tests in Polis check whether a comment's agree/disagree
rate is significantly above 0.5 — a directional hypothesis requiring
one-tailed tests. The two-tailed values were 28% more conservative,
causing fewer comments to pass significance.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: jucor

delphi/polismath/pca_kmeans_rep/repness.py

@@ -15,9 +15,11 @@
 from polismath.utils.general import AGREE, DISAGREE
 
 
-# Statistical constants
-Z_90 = 1.645  # Z-score for 90% confidence
-Z_95 = 1.96   # Z-score for 95% confidence
+# Statistical constants — one-tailed z-scores, matching Clojure (stats.clj)
+# and Python's own stats.py (z_sig_90, z_sig_95).
+# One-tailed: P(Z > z) = α, i.e. the entire rejection region is on one side.
+Z_90 = 1.2816  # Z-score for 90% confidence (one-tailed)
+Z_95 = 1.6449  # Z-score for 95% confidence (one-tailed)
 
 # Pseudocount for Bayesian smoothing (Beta prior)
 #