Outstanding issues with the well_trajectory forward model

[verveerpj]

Following issues have been noted during development and may need to be revisited:

Issues using the ResInsight-based code:

• Error handling is not optimal: errors originating from ResInsight/rips occur in the log, but do not raise an exception.
• Currently the cases that are loaded must be metric.
• The generated trajectories always start at zero depth, even if the platform z-location is non-zero, or if no platform is specified (with a non-zero first guide-point). This may lead to unintended consequences for the unsuspecting user. Currently the platform depth can be optimized, which would not have an effect. See PR Refactor the code for reading trajectories, allowing for missing platforms and kickoffs #27.
• An error is generated by the following call: intersection.update():

  ERROR: Line 1: Expected three dimensions in the vector, got 1
  ERROR: Line 1: Array argument is missing start '['. [[7500.0, 7500.0, 50.0], [7500.0, 7500.0, 8325.0], [7500.05, 7500.0, 8400.0], 
  [7500.1, 7500.0, 8425.0]]" argument of the command: "Assign value to field Points"
  

  It does not seem to cause harm, but it is unclear if there are unforeseen consequences. It also not clear if this always happens or otherwise under which conditions this happens.
• Currently when the trajectories are generated by the built-in simple interpolation, the results can optionally be stored as text files (`PATH_*.txt1). Instead we should write the files in the same standard format as ResInsight, and make it non-optional, as the resinsight code does.
• The current approach to modify the npv input file is not elegant and confusing. Instead of modifying/adding the cost field in the npv input file, it would be beter to add a cost_per_km field, and have an additional well length input. The user then simply provides the npv input file as they are used to and the well trajectory forward model writes the well length file.

[verveerpj]

PR #27 resolves the issue with the platform z-location. It is not possible to set the z-location it is now always assumed to be at zero depth and it cannot be optimized. If platforms are missing they are added at the location straight above the first guide point, without a kickoff.