v0.1.0a17: tbl_one Rao-Scott footnote grade + NaN weight guard

Two fixes from external adversarial audit:

F-02: Rao-Scott rendered table footnote now names the first-order
Kish-DEFF approximation and its known disagreement range with R
survey::svychisq (10-15% or more under high ICC). Previously the
footnote only said "Tests: Rao-Scott chi-square" with no grade
qualifier — invisible in published HTML/PDF tables even though the
UserWarning existed. A separate footnote line is appended whenever
the Rao-Scott test appears in test_names.

F-03: tbl_one() now raises ValueError when the weights column
contains NaN values, matching tbl_regression() behavior. Previously
NaN weights were silently coerced to 0, which caused the affected
group to display N=0 and all statistics as em-dashes with no user
signal. The error message names the column and suggests a fix.

Author: jturner-uofl

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "pysofra"
-version = "0.1.0a16"
+version = "0.1.0a17"
 description = "Statistical reporting and table preparation framework for Python — the missing reporting layer."
 readme = "README.md"
 license = { text = "GPL-3.0-or-later" }

src/pysofra/__init__.py

@@ -50,7 +50,7 @@
 from .summary.tests import available_tests
 from .themes.registry import available_themes, register_theme
 
-__version__ = "0.1.0a16"
+__version__ = "0.1.0a17"
 
 __all__ = [
     "CellPart",

src/pysofra/summary/tbl_one.py

@@ -159,6 +159,14 @@ def tbl_one(
                 "Negative weights are not supported; drop or correct them "
                 "before calling tbl_one()."
             )
+        n_nan = int(w_col.isna().sum())
+        if n_nan > 0:
+            raise ValueError(
+                f"weights column {weights!r} contains {n_nan} NaN value(s). "
+                "Drop or impute missing weights before calling tbl_one(). "
+                "(Use data.dropna(subset=[weights_col]) or fill with a "
+                "suitable value first.)"
+            )
         total = float(w_col.fillna(0.0).sum())
         if total <= 0:
             raise ValueError(
@@ -454,6 +462,12 @@ def _fmt_n(val: float | int) -> str:
         footnotes.append("n (%) for categorical variables.")
     if show_p and test_names:
         footnotes.append("Tests: " + "; ".join(sorted(test_names)) + ".")
+        if "Rao–Scott chi-square" in test_names:
+            footnotes.append(
+                "Rao–Scott chi-square: first-order Kish-DEFF approximation; "
+                "may disagree with R survey::svychisq (second-order) by 10–15 % "
+                "or more under complex designs with strong intra-cluster correlation."
+            )
     if show_q:
         footnotes.append(f"q-value = {_q_method_label(q_method)} adjusted p-value.")
     if show_smd: