Redundant lines: 2 bug fixes + synthetic warning

[Didou09]

Motivations:

• Bugfix to set dict keys to ADAS indexing was not working properly when index_lines=False because:
  • isel was not defined (but used)
  • dict key was set priori to shifting
• Lines with redundant wavelength in a adf15 file seem to be more common than orignally thought and several uneasy cases can occur.

Main proposed changes:

• The 2 bugs are fixed
• A synthetic warning is raised when lines are shifted due to redundancy, it lists all lines and wavelength and makes visible some uneasy cases with mulitple redundancies and/or unfortunate discrepancies exactly equal to the implemented shift by +1e-6 A.
• See here for details and thoughts on possible solutions.

Feel free to merge or just to use as back-up.

[fsciortino]

thanks for catching those bugs @Didou09 ! I'll merge this as it is (I don't seem to be able to push to your fork, so I cannot make changes in this PR, I think...), but then I will also immediately change the 1e-6 factors to 1e-10, just in case.

As you know, changing default behavior for such a function would be a bit of a problem for backwards compatibility... indeed it would probably affect a few users. I would prefer to leave the default as it is now for the moment, but remind people to set their value of index_lines explicitly when they use this function to avoid future issues.

[Didou09]

OK, do you want for future PRs, that I push on a dedicated branch in your repo instead of doing it via my fork ?
This way you would be able to push to said branch.

If that solution is better for you you need to make sure to include me as a collaborator in your settings so I can actually push to your repo direcly, on non-critical branches.

[fsciortino]

just added you as collaborator 👍