Fix travis: suppress catkin_lint error for eigenpy

[rhaschke]

See fkie/catkin_lint#73

[v4hn]

Could you add a TODO to remove this for noetic?

[rhaschke]

This is not related to Noetic. We can remove it if catkin_lint relaxes the check.
If @roehling decides not to relax the check, we need to find_package(eigenpy) on its own and consider the corresponding _INCLUDE_DIRS and _LIBRARIES variables explicitly. Currently, they just become part of the corresponding catkin_* variables.
I primarily filed this to fix our Travis, particularly in view of WMD next week.

[v4hn]

we need to find_package(eigenpy) on its own and consider the corresponding _INCLUDE_DIRS and _LIBRARIES variables explicitly

And that's the way I would like to handle this in the future, as I agree that this should not be a catkin component if it's not built through catkin. Thus the request for the TODO.

Anyway, as we plan to remove the dependency in the future in favor of boost_python anyway, I'm also fine with this as-is until then.

Can be merged as soon as CI succeeds.

[rhaschke]

If you agree that eigenpy shouldn't be a catkin component if it's not building via catkin, we should directly prefer a cleaner (but more laborious) solution as suggested in #1734 (comment).
IMHO, if a package provides a package.xml it should be accepted as a proper ROS package that can be imported via find_package(catkin ...). Let's wait for feedback from @roehling.

[v4hn]

if a package provides a package.xml it should be accepted as a proper ROS package that can be imported via find_package(catkin ...).

I did not look at the actual implementation of the catkin-compatible cmake config.
If it is actually merged (and released) in upstream sources, I would have been fine with your solution too.
The use-case I have in mind is other linux distributions which would package eigenpy as a ROS-agnostic library from upstream sources.