Remove miniature scale from science data and fix segfault

[chapulina]

• Broken off Visualize science data as points #88

Tests were crashing for that PR, while investigating, I noticed it happened because the MINIATURE_SCALE had been removed.

This PR removes the miniature scale and prints a warning instead of segfaulting, but I haven't looked closely into the barycentric interpolation to see why the miniature scale caused the issue, could you take a look @mabelzhang ?

I also added other changes from #88 which help us not need the miniature scale:

• The invisible ground plane makes it possible to easily scroll to a far away distance
• This can be seen with the 300kmx300km grid that is also added here
• The increased camera clipping distances are also needed to see the entire grid from a distance
• Added the camera controls plugin so the user can easily go back to the initial camera angle

[mabelzhang]

Thank you for debugging! Are you able to find data to test it locally? If I just delete the added if-statements, it still doesn't seg fault or hang for me.

I checked the hanging test at https://github.com/osrf/lrauv/pull/88/checks?sha=fa28e234918d0c2a86232d272e931095e2c1600c , and the output

10: [Wrn] [ScienceSensorsSystem.cc:858] 1D linear interpolation: query point lies on one side of all interpolation points. Cannot interpolate.

suggests that it returned NaN and something continued looping, possibly one of the loops in PostUpdate():

  for (auto &[entity, sensor] : this->entitySensorMap)

On a side note, the bad news of looking at this actual data is that, I now remember why we needed the per-slice searches as opposed to barycentric. Ughhhhhh. It's the z resolution being significantly smaller than x and y! The 4 NNs are all in 1 line along z. We might have to go back to trilinear interpolation. Ugh. At least I haven't deleted those branches.

To make our lives easier, we might have to take a step back and make the strict assumption that input indices are sorted, like in netCDF.

On a related note, debugging the interpolation will be significantly easier if we can publish a visual Marker from the ScienceSensorsSystem plugin. Anything against that?

The last thing I left off in the trilinear interpolation branch is that the found 4 NNs didn't satisfy the assumption that they need to be in a rectangular prism, which was the reason I moved to the less restrictive barycentric interpolation. I would publish a Marker to see if there are any bugs to shake out, before giving up and rewinding to assume all the input indices are sorted.

Verdict: real data is ugly.

[mabelzhang]

Should be fixed in b8641f0
So the cause was a pretty important bug - in the 1D case of interpolation, I never stored the array after sorting, only stored the sorted indices, heh.
I'm gonna approve my own fix 😅 Try it and see if the new warnings still print for you - they shouldn't anymore.

[chapulina]

Warnings are gone! Thanks for fixing the bug :D