Incorrect computation of population layouts if alternative clustering is enabled

[ekatef]

It appears that output of build_clustered_population_layouts.py may be not correct if alternative clustering is enabled. Both the overall population and distribution are incorrect.

Checklist

• I am using the current main branch or the latest release. Please indicate.
• I am running on an up-to-date pypsa-earth environment. Update via conda env update -f envs/environment.yaml.

Describe the Bug

Ways to reproduce

Testing for Kazakhstan

Issues found:

• after clustering, population is ~44 mlns (instead of ~20 mlns in reality)
• population of the former Karaganda region (Ulytau and Karaganda regions according to the actual administrative division) is ~16 mlns after clustering instead of 1.3 mlns

Testing for USA

Issues found:

• after clustering, the overall population is ~3 times higher than the real value
• population distribution is not realistic

[ekatef]

@hazemakhalek may you have any ideas on why it could happen?