refactor(waterdata): collapse _coerce_targets to a single scalar guard

The `isinstance(parsed, pd.DatetimeIndex)` early-return was redundant:
`pd.DatetimeIndex(parsed)` returns an equivalent index for an already-
parsed DatetimeIndex, so the three branches reduce to one scalar guard
that wraps a bare value in a list before constructing the index. Behavior
is unchanged for every input kind (DatetimeIndex, scalar string/Timestamp,
list, Series, ndarray, empty) — the existing nearest tests cover them all.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sjb14HkwuCydKSKMsaXsgd

Author: thodson-usgs

dataretrieval/waterdata/nearest.py

@@ -187,10 +187,8 @@ def _coerce_targets(targets: Any) -> pd.DatetimeIndex:
     is wrapped directly so its elements are preserved.
     """
     parsed = pd.to_datetime(targets, utc=True)
-    if isinstance(parsed, pd.DatetimeIndex):
-        return parsed
     if pd.api.types.is_scalar(parsed):
-        return pd.DatetimeIndex([parsed])
+        parsed = [parsed]
     return pd.DatetimeIndex(parsed)