Add 2025 survey data support

The 2025 survey uses a single English-only xlsx (instead of separate
fi/en files) with a restructured schema: compensation is split into
base salary, commission, lomaraha, bonus, and equity components;
working time is h/week instead of percentage; and competitive salary
is categorical instead of boolean. Vuositulot is now synthesized
from the component fields.

Drop COLUMN_MAP_2024, COLUMN_MAP_2024_EN_TO_FI, VALUE_MAP_2024_EN_TO_FI,
read_initial_dfs_2024, read_data_2024, map_sukupuoli, map_vuositulot,
split_boolean_column_to_other, apply_fixups, and the associated gender
value lists and boolean text maps. All of this exists in version history.

- KKPALKKA now includes base salary + commission (median 5500 → 5800)
- Apply map_numberlike to tuntilaskutus and vuosilaskutus columns to
  handle string values like "60 000" and "100 000"
- Filter out zeros when computing tunnusluvut on the index page so
  stats reflect actual reported values

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

pulkka/data_utils.py

@@ -43,3 +43,28 @@ def rename_na(df: pd.DataFrame, col: str, na_name: str) -> None:
     df[col] = df[col].astype("string")
     df.loc[df[col].isna(), col] = na_name
     df[col] = df[col].astype("category")
+
+
+def explode_multiselect(
+    series: pd.Series,
+    *,
+    sep: str = ", ",
+    top_n: int | None = None,
+) -> pd.Series:
+    """
+    Explode a comma-separated multiselect column into value counts.
+
+    Returns a Series of counts indexed by individual values,
+    sorted descending. Optionally limited to top_n entries.
+    """
+    counts = (
+        series.dropna()
+        .str.split(sep)
+        .explode()
+        .str.strip()
+        .loc[lambda s: s != ""]
+        .value_counts()
+    )
+    if top_n is not None:
+        counts = counts.head(top_n)
+    return counts