Fix dataframe time column period formatting

This is a (small) hack that works around changes in how pandas deals
with casting Series to `datetime64` types have units larger than
seconds. We depended on the previous behavior. Not sure if it's
something that will get fixed in pandas, but I made this issue and it's
a breaking change, so hopefully:

pandas-dev/pandas#48574

src/pudl/metadata/constants.py

@@ -2,6 +2,7 @@
 import datetime
 from collections.abc import Callable
 
+import numpy as np
 import pandas as pd
 import pyarrow as pa
 import sqlalchemy as sa
@@ -72,12 +73,12 @@
 """
 
 PERIODS: dict[str, Callable[[pd.Series], pd.Series]] = {
-    "year": lambda x: x.astype("datetime64[Y]"),
+    "year": lambda x: pd.Series(np.array(x).astype("datetime64[Y]")),
     "quarter": lambda x: x.apply(
         pd.tseries.offsets.QuarterBegin(startingMonth=1).rollback
     ),
-    "month": lambda x: x.astype("datetime64[M]"),
-    "date": lambda x: x.astype("datetime64[D]"),
+    "month": lambda x: pd.Series(np.array(x).astype("datetime64[M]")),
+    "date": lambda x: pd.Series(np.array(x).astype("datetime64[D]")),
 }
 """
 Functions converting datetimes to period start times, by time period.