Add absolute tolerance to KMeans algorithm

[MassimoCimmino]

This closes #1985 and #1976.

I did not add any test since the issue occured in the already-implemented cases. Issue #1976 was caused by an error in the initialization of centroid positions.

[mwetter]

@MassimoCimmino : Thanks for the changes. The verification fails due to a change in result of 0.4 K in one model, and in IBPSA.Utilities.Clustering.Validation.KMeans_1d, the returned clustering is different from the expected labelsExp. Can you please review and update as needed.

[MassimoCimmino]

I updated labelsExp in IBPSA.Utilities.Clustering.Validation.KMeans_1d, IBPSA.Utilities.Clustering.Validation.KMeans_2d and IBPSA.Fluid.Geothermal.Borefields.BaseClasses.HeatTransfer.ThermalResponseFactors.Validation.ClusterBoreholes_100boreholes. The changes to IBPSA.Utilities.Clustering.KMeans enables the algorithm to find clusters with lower inertia. The difference in clusters in IBPSA.Fluid.Geothermal.Borefields.BaseClasses.HeatTransfer.ThermalResponseFactors.Validation.ClusterBoreholes_100boreholes would also explain the new results from the other models which use the same borefield geometry. I updated the results there also.