Add auto type casting in io

[DaniilSNikulin]

On PCL all vertices and normals exist only in float32 format. However, a lot of libraries and tools use float64 format. In original version, PCL don't recognize fields if types are not equals.

This commit add auto type casting in conversion from pcl::PCLPointCloud2 to pcl::PointCloud. This is allow read files, which are use float64 format.

To reproduce test case, was written simple unit test.

[taketwo]

Hey, thanks for the effort, this is a very interesting contribution. I personally was hit by this limitation recently when I tried to load a PLY file with f64 fields.

I'm very much in favor of merging this. However, since the implementation is not absolutely trivial, I'd like to have a closer look. I think it's important to make sure that we are not introducing bugs (obviously) and any noticeable time penalty (these conversion functions are used a lot in ROS code, so we will be affecting loooots of people). Unfortunately I don't have enough time to give this PR attention it deserves right now, but will try as soon as possible.

[DaniilSNikulin]

remark from 3337 issue:

During concatenation, a check of field datatype should be performed

Done

I am afraid this declaration from 3337 are not fully correct:

Currenty, PCLPointCloud2::concatenate assumed same name implies same datatype. WIth the PR #3312 this will no longer be true.

Master version of PCL already successfully read float64 fields into PCLPointCloud2, current PR just allow type casting to PointT in function fromPCLPointCloud2 after data was already in PCLPointCloud2. Also PCLPointCloud2 could be obtained from ROS message with any fields type.

This means that all functions in pcl with same hypotheses potentially have errors.

[stale]

This pull request has been automatically marked as stale because it hasn't had
any activity in the past 60 days. Commenting or adding a new commit to the
pull request will revert this.

Come back whenever you have time. We look forward to your contribution.