Improved distance (matrix) computation

[astamm]

• fixes in find_rotation_seed_unique() and find_rotation_seed_coord(); penalty option is added for both;
• refactoring in calc_shape_dist() and curve_dist() using newly defined internal functions curve_to_srvf() and match_f2_to_f1();
• New internal functions in utils-pointwise.R:
  • amplitude_distance() computes the distance in amplitude between two SRVFs supposed optimally transformed;
  • phase_distance() computes the distance in phase between two SRVFs assumed to be already optimally transformed using the estimated $\gamma$;
  • pointwise_mean() computes or updates the Karcher mean or median using a set of SRVFs assumed to be already optimally transformed.

These internal functions should be helpful in making the code for Karcher mean computation clearer in the near future.

• Fixes in kmeans_align() which was broken; still needs an update though; in a near future as well.

[jdtuck]

See code review above

[astamm]

Thanks but I do not see the review.

[jdtuck]

where did the centering go? This is a large PR so have had hard time going through the diff to find it if you moved it.

[astamm]

I have it done under curve_to_srvf() in curve_functions.R.

[jdtuck]

Is this ready to merge?

[jdtuck]

closes #44 and #42

[astamm]

Well it does pass current tests and should not break current user experience and should make one or two things faster.

[astamm]

In one of my commits, I added Closes #42 but I wouldn't close #44 as I have not yet done this.