OPR mode support for PIE type reconstructors

[hruth]

• Removed OPR mode update methods from LSQMLReconstructor and added them into OPRModeWeights
• Removed OPR mode settings from LSQML options classes and added them to the base options classes
• Added OPR mode support for PIE type reconstructors

[mdw771]

Can you also add Ash as a reviewer?

[hruth]

I also found a bug in update_opr_probe_modes_and_weights .

probe[i_opr_mode, 0, :, :] = eigenmode_i updates reconstructor.parameter_group.probe.data since probe is a reference to reconstructor.parameter_group.probe.data. It seems like this is unintentional. The same is true for the weights. You should expect that the fix to this will change the algorithms results, but only for the case where probe optimization is on and weight optimization is off or vice-versa. I'll fix this issue in the next commit.

[hruth]

Can you also add Ash as a reviewer?

I just did. Annoyingly, GitHub only allows a single reviewer at a time if you are using the free version. I suppose it's good enough to just add reviewers so that they get an email notification about it, even if they don't keep the reviewer status on the PR.

[mdw771]

I also found a bug in update_opr_probe_modes_and_weights .

probe[i_opr_mode, 0, :, :] = eigenmode_i updates reconstructor.parameter_group.probe.data since probe is a reference to reconstructor.parameter_group.probe.data. It seems like this is unintentional. The same is true for the weights. You should expect that the fix to this will change the algorithms results, but only for the case where probe optimization is on and weight optimization is off or vice-versa. I'll fix this issue in the next commit.

self.parameter_group.probe.data returns a copy. See ReconstructParameter.data(). The updated copies are set back to the probe and OPR weight objects in the subsequent lines.