Update seafloor gridding for mid-ocean ridges that are topological lines

[jcannon-gplates]

Update seafloor gridding for mid-ocean ridges that are topological lines

See #364 for a description of why this was needed. Closes #364.

This comment was written with Claude's help. However, the code is all-human (that'll soon change though).

Summary

• Switch MOR seed-point generation to calculate_plate_boundary_statistics — replaces
  the old separate_ridge_transform_segments-based approach with PyGPlates 1.0's
  calculate_plate_boundary_statistics. MOR features no longer need pre-computed spreading
  properties (left/right plate IDs in the feature data); spreading rates are now derived
  directly from left- and right-plate velocities relative to the ridge velocity at each
  uniformly-spaced boundary point.

• Support asymmetric spreading — each MOR boundary point produces two seed points: one
  shifted into the left plate and one into the right plate, each carrying its own
  independently computed spreading rate. The old code assumed symmetric spreading (the same
  rate was assigned to both sides).

• Suppress anomalous spreading rates caused by missing rotations near MOR appearance
  time — MOR velocity is calculated over the interval [time + 1, time]. If the MOR
  feature lacks rotations at time + 1 (which can happen when a MOR only has rotations
  spanning its active lifetime), the computed velocity can be spuriously large. A new
  _could_mor_have_anomalous_velocity() helper detects this condition per MOR shared
  sub-segment and sets the spreading rate to NaN for all its points rather than writing bad
  values to the grid.

• Return NaN spreading rate (not drop the point) when no left/right plate exists — if
  a boundary point has no resolved left or right plate (e.g. a gap in plate/network coverage),
  the shifted seed point is still emitted but its spreading rate is set to NaN instead of the
  point being silently discarded. This preserves the point for age-grid reconstruction while
  clearly marking the spreading rate as unknown.

• Shift seed points using the boundary normal — the rotation pole used to displace a
  boundary point into its plate is now computed from the cross product of the boundary point
  and the boundary normal, rather than from the plate velocity vector. The boundary normal is
  always perpendicular to the boundary point and well-defined, so the previous
  is_zero_magnitude() guard and the post-shift topology containment check are no longer
  needed.