fix: replace pandas APIs removed in 2.0 that leave four entry points …

[EnvDroneSense]

Changes proposed in this PR:

• Replace the four calls to DataFrame.append and Series.iteritems, removed in pandas 2.0, that make their enclosing functions raise AttributeError: litpop.py:640, impact_data.py:422, calibration_opt.py:406 and calibration_opt.py:94.
• Add one regression test per site. None of the four lines was previously executed by the test suite.

This PR fixes #1319 and #826

The affected entry points are LitPop.from_shape_and_countries (with a GeoSeries or list shape), impact_data.hit_country_per_hazard, calibration_opt.calib_all, and calibration_opt.calib_instance (with a multi-row df_out and yearly_impact=True). pyproject.toml declares pandas with no version constraint, so a fresh install always resolves to an affected version.

litpop.py:640 is the cause of #826; that issue's traceback names this line, and its reporter proposed the same patch.

The fix differs by site. litpop.py and calibration_opt.py:406 accumulate DataFrames, so they collect into a list and pd.concat once, which also removes the quadratic copy the loops paid. impact_data.py:422 appends a dict rather than a frame, so it collects records and builds one pd.DataFrame(records, columns=[...]). calibration_opt.py:94 is a rename.

Two behaviour details worth review:

• calib_all returns the single frame directly when there is exactly one parameter combination. The old code deep-copied the first df_out and only applied ignore_index=True from the second onward, so a caller-supplied impact_data_source index survived; pd.concat(..., ignore_index=True) alone would silently replace it.
• In litpop.py the type check moved out of the loop, ahead of any selection. Previously [Polygon, Point] ran a full point-in-polygon pass for the polygon before raising on the point.

Also calibration_opt.py:406 passed the builtin input as the second positional argument (ignore_index). bool(input) is True, so it happened to mean ignore_index=True.

Each test was checked by reverting its own fix alone, with the other three in place:

reverted                    failing test                          error
litpop.py                   test_geoseries_shape                  'GeoDataFrame' has no attribute 'append'
impact_data.py              test_hit_country_per_hazard_pass      'DataFrame' has no attribute 'append'
calibration_opt.py:406      test_calib_all_multiple_params        'DataFrame' has no attribute 'append'
calibration_opt.py:94       test_calib_instance_yearly_multirow   'Series' has no attribute 'iteritems'

Note there is precedent for this bug class here: test_emdat_impact_yearlysum_no_futurewarning was added when the same .append problem was fixed in emdat_impact_yearlysum. This brings the remaining four sites in line with that.

PR Author Checklist

• Read the Contribution Guide
• Correct target branch selected (if unsure, select develop)
• Descriptive pull request title added
• Source branch up-to-date with target branch
• Documentation updated
• Tests updated
• Tests passing
• No new linter issues
• Changelog updated

PR Reviewer Checklist

• Read the Contribution Guide
• CLIMADA Reviewer Checklist passed
• Tests passing
• No new linter issues