Set SAS to "stability assist" on reference frame change

[simon-dima]

If SAS is set to follow the prograde/tangent direction (resp. normal, binormal), upon changing reference frames the vessel will reorient to the "new" tangent direction. This can waste time or RCS fuel, and it is especially problematic during manœuvre execution.

I propose that changing the reference frame used to determine the directions of the navball markers should set SAS to the "stability assist" mode (maintaining heading, marked with a ~ symbol) if it was previously one of {±tangent, ±normal, ±binormal}. The ±target and "hold manœuvre" directions do not depend on the reference frame currently selected and can be kept.

Related thoughts:

• The only situation I can think of where the current behavior is useful is launch from a body into orbit (and landings, vice versa). Using the body-fixed frame is useful for the early stages of the ascent, setting SAS to tangential in order to reduce atmospheric drag; approaching orbit, the inertial frame is most appropriate. When the navball switches from surface mode into the body-centered inertial frame, the vessel pitches to track the new "orbital" tangent vector; switching to "stability assist" instead would likely surprise users accustomed to this behavior, and would require an additional user input to reactivate tangent vector tracking.
  • To keep the current behavior in this case, SAS mode could be preserved when the reference frame is changed automatically, setting it to "stability assist" only when the reference frame is changed from the Principia interface.
• If "navball frame" and "plotting frame" become distinct as suggested in Make navball+trajectory plotting frame different combinations possible #2608, SAS should follow the Frenet vectors of the navball frame, and this issue is then about changing the navball frame. The current behaviour of SAS modes persisting on frame change would be less bothersome, since it is problematic mainly when changing the plotting frame from the map view, in my experience.
• The proposed behavior would ensure that the vessel direction targeted by SAS remains continuous, without jumps, when changing the reference frame. The angular velocity required to match the targeted direction may change discontinuously, since the Frenet vectors associated to the vessel's trajectory in the original reference frame are non-constant. The rotational speeds involved are very small, so I think this effect is entirely unproblematic in most situations, but it may be relevant to attitude-keeping (see Make navball+trajectory plotting frame different combinations possible #2608 (comment), Disable timewarp rotation if under a certain threshold #2556).
  • When the selected reference frame rotates (e.g. body-fixed, satellite-body-orbit), does the "stability assist" SAS mode aim to keep the direction vector constant in the selected reference frame or constant in an inertial frame? Both seem like reasonable choices. If it is the first possibility, this would be another source of discontinuity in targeted angular velocity.