Targeting to use Levenberg Marquardt

Currently, the targeter (differential corrector) uses a gradient descent method. This works, but it's slow. An optimization crate provides the LM algorithm out of the box: https://crates.io/crates/levenberg_marquardt .

• Create a targeting instance, similar to PropInstance
• Update framework to allow for quick switching between Levenberg and Newton Gauss
• Update targeting documentation and test cases
• Consider whether to keep gradient descent for posterity, comparison, and verification of targeting method.

NOTE: This will likely cause breaking changes in the Targeting API.