add kwarg to local exceedance intensity to change value ordering

[ValentinGebhart]

Changes proposed in this PR:

• add new kwarg reverse to methods Hazard.local_exceedance_intensity, Hazard.local_return_period, util.interpolation.preprocess_and_interpolate_ev such that the user can decide how to order intensities, for instance when handling negative intensities like the SPIE dorught index.
• added short explanation to climada_util_local_exceedance_values.ipynb

This PR addresses #1257.

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

[ValentinGebhart]

Example usage:
1) Typical hazard object with positive values

client = Client()
haz_tc_haiti = client.get_hazard(
    "tropical_cyclone",
    properties={
        "country_name": "Haiti",
        "climate_scenario": "historical",
        "nb_synth_tracks": "10",
    },
)
local_exceedance_intensity, title, column_label = (
    haz_tc_haiti.local_exceedance_intensity(return_periods=[50, 200])
)
axes = plot_from_gdf(local_exceedance_intensity, title, column_label, figsize=(8, 4))

2. Hazard with negative values but usual ordering (e.g., Temperatures in degrees Celsius)

shift_intensity = -100.0
haz_neg_usual_order = copy.deepcopy(haz_tc_haiti)
haz_neg_usual_order.intensity = sparse.csr_matrix(
    haz_tc_haiti.intensity.toarray() + shift_intensity
)

local_exceedance_intensity, title, column_label = (
    haz_neg_usual_order.local_exceedance_intensity(
        return_periods=[50, 200],
        min_intensity=-1e10,
        log_intensity=False,
        log_frequency=False,
    )
)
axes = plot_from_gdf(local_exceedance_intensity, title, column_label, figsize=(8, 4))

3. Hazard with negative values and inverse ordering (e.g., more negative is more serious, like the SPIE drought index)

haz_neg_inverse_order = copy.deepcopy(haz_tc_haiti)
haz_neg_usual_order.intensity = -1 * haz_tc_haiti.intensity
local_exceedance_intensity, title, column_label = (
    haz_neg_usual_order.local_exceedance_intensity(
        return_periods=[50, 200],
        min_intensity=-1e10,
        log_intensity=False,
        log_frequency=False,
        reverse=True,
    )
)
axes = plot_from_gdf(
    local_exceedance_intensity, title, column_label, figsize=(8, 4), cmap="viridis_r"
)

[luseverin]

I've had a look and it looks sensible to me. The examples added in the notebook are helpful. I would also add in the Notes section of the docstrings of the functions when you are supposed to set reverse=True, e.g.

Use reverse=True when smaller hazard intensities correspond to higher impacts, e.g. in the case of negative winter temperatures or drought indices such as SPEI.

or something similar. Otherwise it looks good to me.

[emanuel-schmid]

there is only one test with reverse=False parameter and this one fails. can we have some more tests?

[emanuel-schmid]

many thanks! and sorry for the delayed review.

[emanuel-schmid]

looks dangerous to me. shouldn't there be a reverse case handling?

while we're at it - wouldn't it even be preferable to downright raise an error here instead of writing warnings to a log that might never be read?

[emanuel-schmid]

what would happen if the complete reverse case handling was omitted below and sorted_idxs were reversed instead?