[tools] Try to fix bloat check when comment has empty cell (#26479)

The bloat check has been failing for months now because
gh_report.py script seems to assume that the first row
in a comment to the analyzed PR has no empty cells, and
for the cc32xx platform, which happens to finish the build
as the first one, a row with a blank section name is added.

I don't know what is the reason of the blank section, but
the script should not give up just because of one invalid
entry as we miss important code size increase warnings.

Also, replace deprecated DataFrame.append with concat().

scripts/tools/memory/gh_report.py

@@ -375,7 +375,8 @@ def merge(df: pd.DataFrame, comment) -> pd.DataFrame:
                     cols, rows = memdf.util.markdown.read_hierified(body)
                     break
         logging.debug('REC: read %d rows', len(rows))
-        df = df.append(pd.DataFrame(data=rows, columns=cols).astype(df.dtypes))
+        df = pd.concat([df, pd.DataFrame(data=rows, columns=cols).astype(df.dtypes)],
+                       ignore_index=True)
         return df.sort_values(
             by=['platform', 'target', 'config', 'section']).drop_duplicates()
 

scripts/tools/memory/memdf/util/markdown.py

@@ -34,7 +34,7 @@ def read_hierified(f):
         for i in range(0, len(header)):
             column = columns[i + 1].strip()
             if not column:
-                column = rows[-1][i]
+                column = rows[-1][i] if rows else '(blank)'
             row.append(column)
         rows.append(tuple(row))