perf(utils): skip recursion into non-container list elements

In convert_to_base64, when iterating list/tuple elements, only recurse
into dicts, lists, and tuples. Strings and numbers can never contain
numpy arrays, so recursing into them wastes ~500K function calls on
figures with large text arrays.

Author: KRRT7

_plotly_utils/utils.py

@@ -126,7 +126,8 @@ def _convert_to_base64(obj, np):
                 _convert_to_base64(value, np)
     elif isinstance(obj, (list, tuple)):
         for value in obj:
-            _convert_to_base64(value, np)
+            if isinstance(value, (dict, list, tuple)):
+                _convert_to_base64(value, np)
 
 
 def cumsum(x):