TimeSeries and TimeSeriesEvent constructor should accept an object to construct from

[felixchenier]

At the moment, we can only construct an empty TimeSeries:

ts = TimeSeries()

or a TimeSeries with given attributes:

ts = TimeSeries(time=time, data=data, ...)

We can't construct a TimeSeries based on another object:

ts = TimeSeries(array)
ts = TimeSeries(dataframe)  <-- we can using ts = TimeSeries.from_dataframe(dataframe)
ts = TimeSeries(list)
ts = TimeSeries(series)
ts = TimeSeries(dict_of_arrays)

All these contructs should be accepted and lead to a TimeSeries, with a matched time attribute, which will allow to typecast lots of types to TimeSeries and therefore will simplify compatibility with other Python tools.

breaking compatibility

This would lead to a compatibility break in some (maybe fictitious) cases where someone initialized a TimeSeries using a list of arguments rather than keyword arguments, because the init signature would move from:

def __init__(
        self,
        time: np.ndarray = np.array([]),
        time_info: dict[str, Any] = {"Unit": "s"},
        data: dict[str, np.ndarray] = {},
        data_info: dict[str, dict[str, Any]] = {},
        events: list[TimeSeriesEvent] = [],
    )

to:

def __init__(
        self,
        source: TimeSeries | pd.DataFrame | pd.Series | ArrayLike,
        *,
        time: np.ndarray = np.array([]),
        time_info: dict[str, Any] = {"Unit": "s"},
        data: dict[str, np.ndarray] = {},
        data_info: dict[str, dict[str, Any]] = {},
        events: list[TimeSeriesEvent] = [],
    )

I think this is easy to solve in existing code, and it brings a lot of value.