refactor(waterdata): module coherence — unify id extraction and empty-frame idiom

Two behavior-preserving coherence fixes in the OGC response path (live
get_daily/get_continuous/get_monitoring_locations bit-for-bit unchanged; 184
tests + README/time-conventions doctests pass):

- `_get_resp_data`'s geopandas branch now extracts `id` the same way as its own
  non-geopandas branch — `[f.get("id") for f in features]` — instead of a
  second full `pd.json_normalize(features)["id"]`. Drops a redundant
  whole-feature normalize pass and is None-safe (the json_normalize form
  KeyErrors on id-less features); the two halves of the function now match.
- Add `_empty_frame(geopd)` and call it at the four sites that spelled out
  `gpd.GeoDataFrame() if geopd else pd.DataFrame()`, centralizing the
  "empty page, preserve the geo/plain type" invariant that was folklore
  repeated across the OGC and stats parsers.

Deferred as a coherence backlog (out of scope for this bugfix PR — mostly
public-API or behavior changes): unifying error handling across the
samples/codes/ratings direct-httpx callers (raw raise_for_status vs typed
_raise_for_non_200), the ssl_check exposure asymmetry, get_codes returning a
bare DataFrame, get_stats_data's inline finalizer vs the OGC closure, the
samples GET-tail duplication, and assorted api.py docstring/naming drift.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: thodson-usgs

dataretrieval/waterdata/utils.py

@@ -722,6 +722,17 @@ def _next_req_url(
     return None
 
 
+def _empty_frame(geopd: bool) -> pd.DataFrame:
+    """Empty result frame that preserves the result type.
+
+    Returns a ``GeoDataFrame`` when geopandas is installed (so a later
+    ``pd.concat`` with a real geo page keeps geometry/CRS) and a plain
+    ``DataFrame`` otherwise. Centralizes the "empty page, keep the type"
+    invariant the OGC and stats parsers both rely on.
+    """
+    return gpd.GeoDataFrame() if geopd else pd.DataFrame()
+
+
 def _get_resp_data(
     resp: httpx.Response,
     geopd: bool,
@@ -768,11 +779,9 @@ def _get_resp_data(
     if body is None:
         body = resp.json()
     if not body.get("numberReturned"):
-        # Preserve the GeoDataFrame type on empty short-circuit so a
-        # downstream ``pd.concat([empty_page, geo_page])`` doesn't
-        # downgrade the geopd-installed user's result to a plain
-        # DataFrame (stripping geometry/CRS).
-        return gpd.GeoDataFrame() if geopd else pd.DataFrame()
+        # No rows on this page — short-circuit with a type-preserving
+        # empty frame (see ``_empty_frame``).
+        return _empty_frame(geopd)
 
     # Defensive: a 200 with ``numberReturned > 0`` but missing
     # ``features`` is a real schema-drift shape (mirrors the guard in
@@ -781,7 +790,7 @@ def _get_resp_data(
     # transient transport error to ``_paginate``'s exception handler.
     features = body.get("features") or []
     if not features:
-        return gpd.GeoDataFrame() if geopd else pd.DataFrame()
+        return _empty_frame(geopd)
 
     if not geopd:
         df = pd.json_normalize([f.get("properties") or {} for f in features], sep="_")
@@ -798,7 +807,9 @@ def _get_resp_data(
 
     # Organize json into geodataframe and make sure id column comes along.
     df = gpd.GeoDataFrame.from_features(features)
-    df["id"] = pd.json_normalize(features)["id"].values
+    # Same id extraction as the non-geopandas branch above: None-safe and
+    # avoids a second full ``json_normalize`` pass over every feature.
+    df["id"] = [f.get("id") for f in features]
     df = df[["id"] + [col for col in df.columns if col != "id"]]
 
     # If no geometry present, then return pandas dataframe. A geodataframe
@@ -1361,7 +1372,7 @@ def _handle_stats_nesting(
     fields like ``geometry.type`` from leaking into the result.
     """
     if body is None:
-        return gpd.GeoDataFrame() if geopd else pd.DataFrame()
+        return _empty_frame(geopd)
 
     # An empty (or missing) features list — a real mid-pagination
     # shape — would otherwise crash the downstream merge with
@@ -1372,7 +1383,7 @@ def _handle_stats_nesting(
     # plain DataFrame and strip geometry/CRS.
     features = body.get("features") or []
     if not features:
-        return gpd.GeoDataFrame() if geopd else pd.DataFrame()
+        return _empty_frame(geopd)
 
     # The geopd-missing warning is emitted once at import (see top of module);
     # doing it here would log per page.