API: read_json should follow general default of microseconds

[jorisvandenbossche]

pd.read_json still returns nanoseconds always:

>>> df = pd.DataFrame({"col": range(5), "dates": pd.date_range("2025", periods=5)})
>>> s = df.to_json(date_format="iso")
>>> pd.read_json(io.StringIO(s), convert_dates=["dates"]).dtypes
col               int64
dates    datetime64[ns]
dtype: object

Since JSON is a format like CSV that has no type / unit information stored, we should follow the general default of microseconds (or nanoseconds if those are present), like we do for CSV (or HTML, XML, excel, etc)