Bump CMake minimum version to 3.1.

[SergioRAgostinho]

I know this is already being addressed in #2382, but I want to start dealing with the policies and it might take some time until that other one is merged. Activating the C++14 flags will still be the responsibility of the original PR.

Cheers

[SergioRAgostinho]

There might be an issue with linking to downstream targets. Still confirming.

[taketwo]

I was thinking about the CMakeLists in tutorials, do they need to be updated as well? My conclusion was no, not yet. It should still be possible to use PCL in downstream projects with old CMake. But once we switch to C++14 we should consider bumping version there and requesting CXX_14 explicitly.

[SergioRAgostinho]

I was thinking about the CMakeLists in tutorials, do they need to be updated as well?

That's an interesting point. Since our tutorials are not versioned, it should reflect the current official state of the library. All tutorials related to compiling the library definitely need to be update because you wouldn't be able to compile with anything less than CMake 3.1. The other tutorials which define downstream targets using PCL, those should be ok I guess.

But once we switch to C++14 we should consider bumping version there and requesting CXX_14 explicitly.

I'll create a new issue from this one and add it to the C++14 github project.

[taketwo]

Another question I had is whether we should immediately go for 3.5.1. I think we can officially state that the baseline system from now on is Ubuntu 16.04. Why switching to 3.1 now and then again to 3.5.1 later on if we can do this now?

[SergioRAgostinho]

Just confirmed and there's no issues with downstream targets. I'll proceed to update the compilation tutorials

Why switching to 3.1 now and then again to 3.5.1 later on if we can do this now?

To try to keep version as low as possible until we need a specific feature. 3.5.1 should just be seen as the maximum we're willing to go as of now. I know we have the whole "CUDA as a CMake language" topic that will require a further bump, but until then I would stick to 3.1. In some sense this PR will become the blueprint of what needs to be done in the later version bump. Hopefully I will just blindly replicate what I've done here.