Use string values instead of ints to represent axes

It's clearer and more explicit.

ica/analyzers/message_totals.py

@@ -46,7 +46,7 @@ def get_sums_by_day(dfs: ica.DataFrameNamespace) -> pd.DataFrame:
         )
         .resample("D", on="datetime")
         .sum()
-        .rename_axis("date", axis=0)
+        .rename_axis("date", axis="index")
         .assign(is_from_them=assign_lambda(lambda df: df["text"] - df["is_from_me"]))
     )
 

ica/analyzers/totals_by_day.py

@@ -30,7 +30,7 @@ def main() -> None:
             )
             .resample("D", on="datetime")
             .sum()
-            .rename_axis("date", axis=0)
+            .rename_axis("date", axis="index")
             # Filter out any rows for dates where neither person sent a message
             .pipe(pipe_lambda(lambda df: df[df["text"] != 0]))
             # Add a column for the by-day number of messages from the other

ica/core.py

@@ -148,7 +148,7 @@ def get_messages_dataframe(
         )
         # Remove 'attributedBody' column now that it has been merged into the
         # 'text' column
-        .drop("attributedBody", axis=1)
+        .drop("attributedBody", axis="columns")
         # Use a regex-based heuristic to determine which messages are reactions
         .assign(
             is_reaction=assign_lambda(
@@ -268,7 +268,7 @@ def prepare_df_for_output(df: pd.DataFrame) -> pd.DataFrame:
             },
         )
         # Prettify index column name
-        .rename_axis(prettify_header_name(df.index.name), axis=0)
+        .rename_axis(prettify_header_name(df.index.name), axis="index")
         # Make all indices start from 1 instead of 0, but only if the index is
         # the default (rather than a custom column)
         .pipe(