The code was using the "at" method of pandas DataFrame with multiple
indices, and that was accepted by pandas < 1.4.0. But their docs do
say that "at" is for single element access only, whereas "loc" is for
access to multiple elements.

By downgrading the version of the library. See
<https://github.com/flakehell/flakehell/issues/22>.

And some refactoring in metrics code.

Summary of main changes below:

- New metrics cal methods:
    - correlation_aware_euclidean (the new default)
    - haversine_plus_euclidean
    - haversine_plus_manhattan (the only one implemented previously)
- Add "unclusterable_data_columns" general config option
- Allow choice of HDBSCAN's cluster_selection_method
- Metrics has its own section in config file
- Support to weights for dists along proj x and y
- Used and visualised proj params more compatible
- Remove unused "tstep" from domain configs
- Fix a few crashes in outlier removal methods (addresses issue #5)
- Fix a few warnings

haversine --> haversine_plus_manhattan
correlation --> correlation_aware_euclidean

Used, instead of the "geodist" weight, when setting
custom_metrics_method = "correlation", as we calculate distances using
projected coordinates in this case.

To get rid of TBB version warnings. See
<https://numba.discourse.group/t/suppress-tbb-warning/80> and
<http://numba.pydata.org/numba-doc/latest/user/threading-layer.htm>.

Fixes issue #5

Replacing the hardcoded "skip" lists. This is clearer and more useful.

Which takes into account correlations between quantities.
Not yet available for users to choose via config file.