fix(waterdata): numpy array / Series numeric params leaked their repr into the query

[thodson-usgs]

Problem

A numeric param in _NO_NORMALIZE_PARAMS (water_year, year, month, day, thresholds, …) passed as a numpy array or pandas Series fell into the args[k] = v passthrough in _get_args without being materialized to a list. Downstream, both the GET comma-join (",".join(...) if isinstance(v, (list, tuple))) and the chunker (_extract_axes) test list/tuple — so an ndarray/Series was neither comma-joined nor chunked, and its repr leaked into the URL:

get_peaks(water_year=np.array([2020, 2021]))   # -> water_year=%5B2020+2021%5D   (str of the array!)
get_peaks(water_year=pd.Series([2020, 2021]))  # -> water_year=0    2020\n1 ...  (Series repr)
get_peaks(water_year=[2020, 2021])             # -> water_year=2020,2021         (already fine)

Fix

Split the branch so _NO_NORMALIZE_PARAMS values keep their element types (no string-normalization — they're numbers) but a non-string iterable is still materialized to a list, so it comma-joins and chunks like a plain list.

Verification

numpy array  -> water_year=2020%2C2021
pandas Series-> water_year=2020%2C2021
plain list   -> water_year=2020%2C2021   (unchanged)
scalar int   -> water_year=2020          (unchanged)

Added a regression test (test_get_args_materializes_numpy_and_series_numeric_params); existing numeric/construct tests pass; ruff clean.

🤖 Generated with Claude Code