Refactor GDAL installation to micromamba (conda-forge) and add xcube tests

[suriyahgit]

PR Description

This PR improves the stability and reproducibility of the environment used in CI by moving GDAL and related scientific dependencies away from apt-based installation and into a controlled micromamba environment sourced from conda-forge.

Key changes

• Removed GDAL and GIS dependency installation via apt
• Added GDAL installation within micromamba (gdal=3.11.*, strict channel priority)
• Updated CI workflow to mimic the dependency handling approach used in the openEO backend
• Added xcube-related tests under the tests folder
• Ensures compatibility with Python 3.10, 3.11 and 3.12

Motivation

apt GDAL packages are tied to OS release cycles and caused breakages when testing with newer Python versions (e.g., Python 3.12 + GDAL 3.6.2). Installing GDAL from conda-forge provides stable builds and avoids needing to update Docker/CI configurations frequently.

[suriyahgit]

Poetry was implicitly resolving gdal from PyPI and upgrading it to 3.12.0.post1, which conflicted with libgdal 3.11.* (gdal=3.11.*) installed via micromamba in CI, causing build failures. Explicitly pinning:

gdal = ">=3.11,<3.12"

ensures compatibility with the conda-forge GDAL version and prevents accidental resolver upgrades.

[ValentinaHutter]

Would it be possible for you to re-trigger the tests?
In my setup, I cannot install gdal 3.11 anymore, since 3.12 is already released and 3.11 is no longer available. This will happen with 3.12 as soon as 3.13 is released as well, so we cannot really pin gdal to a specific version.