refactor: Better sort

[12rambau]

This PR simply changes the implementation of the sorting algorithm. Chaining sort algo is possible in GEE and I guess it's not less performant than running our own bubble sort using computed indices. The only trick is to run through the properties in reverse order as the last one to run has priority over the others.

I didn't do performances checks yet but my guess is that it's equivalent. readability being much easier.

[Copilot]

Pull Request Overview

This PR refactors the sortMany method in the ImageCollection class to use a simpler chaining approach with Google Earth Engine's native sort functionality instead of a custom bubble sort implementation using computed indices.

• Replaces complex position computation logic with iterative chaining of GEE's native sort operations
• Adds validation to ensure properties and ascending arrays are the same size
• Simplifies the implementation while maintaining the same functionality for multi-property sorting

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 1 comment.

File	Description
geetools/ee_image_collection.py	Main implementation change - refactored sortMany method to use chained sorting
tests/test_ImageCollection.py	Updated test to validate error handling for mismatched array sizes
Multiple serialized test files	Updated expected outputs reflecting the new implementation structure

Comments suppressed due to low confidence (1)

tests/test_ImageCollection.py:525

• The test only validates one specific error case (mismatched array sizes), but doesn't test the actual sorting functionality with the new implementation. Consider adding a test that verifies the sorting behavior still works correctly.

        with pytest.raises(ValueError):

[12rambau]

I made some tests to benchmark this new solution agains the previous one using the following code:

import ee
import geetools
import pandas as pd

ee.Initialize(project="ee-geetools")

with ee.geetools.Profiler() as p:
    ic = ee.ImageCollection("NOAA/GFS0P25")
    icSorted = (
        ic.limit(500)
        .geetools.sortMany(["forecast_time", "creation_time"])
        .limit(30)
    )
    info = icSorted.toList(icSorted.size()).map(lambda x: ee.Dictionary({
        "index": ee.Image(x).get("system:index"),
        "forecast_time": ee.Date(ee.Image(x).get("forecast_time")).format(),
        "creation_time": ee.Date(ee.Image(x).get("creation_time")).format()
    })).getInfo()

df = pd.DataFrame(p.profile)
df["EECU-s"].sum()

they are in competing range:

	old implementation	new implementation
EECU	0.144	0.12

So I guess we are not making computation explode with this modifications.