[BUG] Fix Future warnings with Pandas 2.1.0

[raoulcollenteur]

Describe the bug

Pastas is not working well with Pandas 2.1.0 for now, as shown by errors on GH Actions.

To Reproduce

https://github.com/pastas/pastas/actions/runs/6069992486/job/16465234055?pr=626

Expected behavior

Work.

Python package version

pandas >= 2.1.0

[raoulcollenteur]

Limiting Pandas to 2.0 for now through PR #626, so GH Actions works again and we can work on Pandas 2.1.0 support from there in a separate branch.

[raoulcollenteur]

I checked an I think it might actually work, but we get lots of future warnings from Pandas. So we have to update some of the code. After upgrading the master-branch to 1.3, we can allow the dev-branch to allow Pandas >=2.1 and solve the warnings efficiently.

[raoulcollenteur]

Some warnings I found

• /python3.11/site-packages/pastas/io/pas.py:40: FutureWarning: Passing literal json to 'read_json' is deprecated and will be removed in a future version. To read from a literal string, wrap it in a 'StringIO' object.
  obj[key] = read_json(value, typ="series", orient="split")

• /python3.11/site-packages/pastas/model.py:1179: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.
  parameters = concat(frames)

[martinvonk]

• /pastas/pastas/timeseries_utils.py:135: FutureWarning: Day.delta is deprecated and will be removed in a future version. Use pd.Timedelta(obj) instead
  dt = to_offset(freq).delta / Timedelta(1, "D")

• pastas/pastas/stats/core.py:198: FutureWarning: The 'unit' keyword in TimedeltaIndex construction is deprecated and will be removed in a future version. Use pd.to_timedelta instead.
  index=TimedeltaIndex(lags, unit="D", name="Lags"),

• pastas/pastas/model.py:1801: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '['±12.78%' '±13.98%' '±0.00e+00%' '±0.00e+00%' '±27.49%' '±11.42%'
  '±12.01%' '±0.00e+00%' '±4.27e-02%' '±26.32%' '±14.90%']' has dtype incompatible with float64, please explicitly cast to a compatible dtype first.
  parameters.loc[:, "stderr"] = "±" + stderr.abs().apply(

• pastas/pastas/model.py:1209: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.
  parameters = concat(frames)

• pastas/pastas/model.py:1215: FutureWarning: A value is trying to be set on a copy of a DataFrame or Series through chained assignment using an inplace method.
  The behavior will change in pandas 3.0. This inplace method will never work because the intermediate object on which we are setting values always behaves as a copy.

  For example, when doing 'df[col].method(value, inplace=True)', try using 'df.method({col: value}, inplace=True)' or df[col] = df[col].method(value) instead, to perform the operation inplace on the original object.

  parameters.optimal.update(self.parameters.optimal)

• pastas/pastas/timeseries_utils.py:181: FutureWarning: Day.is_anchored is deprecated and will be removed in a future version, please use False instead.
  if to_offset(offset.rule_code).is_anchored():

[raoulcollenteur]

Pastas now works with all Pandas versions up to the latest 2.2 release after PR #684.

For pandas 3.0, some additional changes are necessary and only pandas 2.2 or higher will work. https://pandas.pydata.org/docs/whatsnew/v2.2.0.html#upcoming-changes-in-pandas-3-0