Make conda install climada the default installation instruction

[peanutfun]

Climada is now available via conda-forge: https://anaconda.org/conda-forge/climada

This means that the tedious dependency installation before running pip install climada can now be dropped in favor of the much simpler conda install -c conda-forge climada. To get to the same result as with the current simple installation instructions, one now has to execute a single command:

conda create -n climada_env -c conda-forge climada

This will install Climada and all its dependencies into a new climada_env environment.

Proposals:

• Add conda install -c conda-forge climada as installation instruction in the README
• Rewrite the Simple Instructions to simply use conda create -n climada_env -c conda-forge climada

[emanuel-schmid]

👍 Very nice, indeed!
But before we change the default installation instruction we have to tackle some dependency issues I'm afraid:

• xarray cfgrib backend changed data variable name #678
• climada.hazard.test.test_tc_tracks.TestIO.test_hdf5_io fails
• climada.hazard.test.test_tc_tracks.TestIO.test_to_geodataframe_line fails but can be fixed with commit fa75338
• climada.engine.unsequa.test.test_unsequa.TestOutput.test_plot_unc_cb fails
• climada.test.test_litpop_integr.TestAdmin1.test_brandenburg and TestLitPopExposure.test_netherlands150_pass fail

[emanuel-schmid]

PR #682 fixes all failing tests with the exception of those related to #678 .

[emanuel-schmid]

Maintenance of the conda-forge installation may turn out to be somewhat demanding.

We'll need to make restrictions in the dependency versions, for domain reasons but also for technical reasons: on my computer I am not able to run python build_locally.py successfully without such restrictions: it takes an unknown amount of time (>5h) to eventually crash, possibly for memory reasons.
Even with most dependency versions fixed, it still requires significant amounts of time (2.5h), cpu (1 core, 100%) and memory (up to 6G).

[peanutfun]

python build_locally.py is not required. The conda-forge CI pipeline automatically tries to build all PRs submitted to https://github.com/conda-forge/climada-feedstock

[peanutfun]

Resolved by #714