Improve calibration methods in HumanStateProvider to use multiple data points

Until now the calibration methods of the `HumanStateProvider` called [here](https://github.com/robotology/human-dynamics-estimation/blob/f55e70d7adebc66552f4359f62b90fbed8b8d7ae/devices/HumanStateProvider/HumanStateProvider.cpp#L315-L384) are using a single data point to compute the calibration matrices. This could be improved by using multiple data points.

Start in devices/HumanStateProvider/HumanStateProvider.cpp at the calibration methods around lines 315-384. Read how the current calibration matrices are computed from one data point, then determine how multiple data points should be incorporated; the work is done when those methods use multiple points for calibration.

Written by the indexing model from the issue text.