Fix CategoricalNB predict's feature-count check and min_categories option

check_dim compared x's feature count against
Nx.axis_size(model.feature_count, 1), but feature_count here is
{num_features, num_classes, num_categories} (unlike the 2D
{num_classes, num_features} in the other NaiveBayes modules this was
copied from), so axis 1 is num_classes. Every predict/*_probability
call rejected correctly-shaped input whenever num_features didn't
happen to equal num_classes, and would silently accept wrongly-shaped
input in the reverse case. Fixed to read axis 0.

Separately, fit's min_categories option was dead: Keyword.pop removed
:min_categories from opts and rebound it, so the very next read of
opts[:min_categories] (used to size the feature_count/
feature_log_probability tensors) always saw nil and silently fell back
to inferring the category count from the training data alone. Kept a
flag captured before the pop and branched on it instead, matching
sklearn.naive_bayes.CategoricalNB's min_categories semantics (a literal
per-feature category count, not a max index) - verified shapes match
sklearn exactly for both list and tensor min_categories inputs.

Author: RicardoSantos-99

lib/scholar/naive_bayes/categorical.ex

@@ -272,10 +272,11 @@ defmodule Scholar.NaiveBayes.Categorical do
     end
 
     num_categories =
-      (opts[:min_categories] || x)
-      |> Nx.reduce_max()
-      |> Nx.add(1)
-      |> Nx.to_number()
+      if min_categories_flag do
+        min_categories |> Nx.reduce_max() |> Nx.to_number() |> trunc()
+      else
+        x |> Nx.reduce_max() |> Nx.add(1) |> Nx.to_number() |> trunc()
+      end
 
     opts =
       opts ++
@@ -360,7 +361,7 @@ defmodule Scholar.NaiveBayes.Categorical do
   """
 
   defn predict(%__MODULE__{} = model, x, classes) do
-    check_dim(x, Nx.axis_size(model.feature_count, 1))
+    check_dim(x, Nx.axis_size(model.feature_count, 0))
 
     if Nx.rank(classes) != 1 do
       raise ArgumentError,
@@ -401,7 +402,7 @@ defmodule Scholar.NaiveBayes.Categorical do
   """
 
   defn predict_log_probability(%__MODULE__{} = model, x) do
-    check_dim(x, Nx.axis_size(model.feature_count, 1))
+    check_dim(x, Nx.axis_size(model.feature_count, 0))
     jll = joint_log_likelihood(model, x)
 
     log_proba_x =
@@ -453,7 +454,7 @@ defmodule Scholar.NaiveBayes.Categorical do
   """
 
   defn predict_joint_log_probability(%__MODULE__{} = model, x) do
-    check_dim(x, Nx.axis_size(model.feature_count, 1))
+    check_dim(x, Nx.axis_size(model.feature_count, 0))
     joint_log_likelihood(model, x)
   end