switch from bool to int for sparse label dtype

avoids FutureWarnings when creating Pandas dfs from scipy sparse matrix types, but also seems like unnecessary memory use, consider reverting to bool pending info on this issue: pandas-dev/pandas#59739

Author: sammlapp

opensoundscape/annotations.py

@@ -1171,19 +1171,22 @@ def integer_to_multi_hot(labels, n_classes, sparse=False):
         labels: list of lists of integer labels, eg [[0,1,2],[3]]
         n_classes: number of classes
     Returns:
-        2d np.array with False for absent and True for present
+        if sparse is False: 2d np.array with False for absent and True for present
+        if sparse is True: scipy.sparse.csr_matrix with 0 for absent and 1 for present
     """
+    # TODO: consider using bool rather than int dtype, much smaller and int is unnecessary
+    # but bool leads to FutureWarning, see https://github.com/pandas-dev/pandas/issues/59739
     if sparse:
         vals = []
         rows = []
         cols = []
         for i, row in enumerate(labels):
             for col in row:
-                vals.append(True)
+                vals.append(1)
                 rows.append(i)
                 cols.append(col)
         return scipy.sparse.csr_matrix(
-            (vals, (rows, cols)), shape=(len(labels), n_classes), dtype=bool
+            (vals, (rows, cols)), shape=(len(labels), n_classes), dtype=int
         )
     else:
         multi_hot = np.zeros((len(labels), n_classes), dtype=bool)
@@ -1213,17 +1216,19 @@ def categorical_to_multi_hot(labels, classes=None, sparse=False):
     rows = []
     cols = []
 
+    # TODO: consider using bool rather than int dtype, much smaller and int is unnecessary
+    # but bool leads to FutureWarning, see https://github.com/pandas-dev/pandas/issues/59739
     def add_labels(i, labels):
         for label in labels:
             if label in classes:
-                vals.append(True)
+                vals.append(1)
                 rows.append(i)
                 cols.append(label_idx_dict[label])
 
     [add_labels(i, l) for i, l in enumerate(labels)]
 
     multi_hot = scipy.sparse.csr_matrix(
-        (vals, (rows, cols)), shape=(len(labels), len(classes)), dtype=bool
+        (vals, (rows, cols)), shape=(len(labels), len(classes)), dtype=int
     )
 
     if sparse:
@@ -1394,10 +1399,6 @@ def find_overlapping_idxs_in_clip_df(
 
 
 from itertools import chain
-from opensoundscape.annotations import (
-    multi_hot_to_categorical,
-    categorical_to_multi_hot,
-)
 
 
 class CategoricalLabels: