Nimblephysics

This release attempts to resolve pelvis angle ambiguity by adding an option to minimize the energy stored in virtual springs attached to the joints coming out of the pelvis.

There are several stability improvements in this release, but the primary improvement is that MarkerFitter now has `fineTuneWithIMU()`, which can optimize a motion to better match IMU data.

Currently when processing very long trials without arms (as show up in the Carmago dataset), if you forget to flag the fact that your model has no arms in the GUI, the optimizer will valiantly attempt to keep marking frames as "missing GRF" until it gets a reasonable solve, but that takes many many hours to eventually decide it's impossible and fail, and stalls the dev server.

This patch with automatically enable the reaction-wheels method of angular residual reduction if it detects the angular fit failing.

This patch scales the weight of the `scaleGRF` regularization term with the length of the trial. This fixes issues for extremely long trials (like the Carmago dataset), where we were getting crazy low masses because the scaleGRF term would take on half the burden.

This release fixes a bug introduced in the previous release, where we did not check if the smart pointer containing the anthropometric prior is `nullptr` before adding it to the "scale + fit" step in `MarkerFitter`.

This release adds an anthropometric prior to the initial IK+scaling steps in MarkerFitter, and makes a number of minor memory safety defensive changes.