rgbd: restore fetching colors for pointcloud in coloredkinfu

[PeterBowman]

Cherry-picked from 04cafeb (authored by @paroj). As pointed out by @alalek in #3121 (comment), this 4.x patch never reached 5.x because of merge conflicts.

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake

[fallenmi]

This replaces a released public pure-virtual method rather than extending it, which breaks source compatibility for external ColoredKinFu implementations. OpenCV 5.0.0 and this PR's exact base expose getCloud(OutputArray, OutputArray) const. A minimal derived class that implements that method with override compiles against both headers; against exact head 6154ffca6036bdf16b5fd06b5049ea3c50fbe507 it fails because the only virtual now has three parameters. The default argument preserves ordinary two-argument calls, but it cannot preserve an override's signature.

Please keep the released two-output virtual intact and add the color-producing API additively (an overload or a separately named method), then add a compile-time compatibility probe for the existing derived-class signature. A controlled old-header/new-library vtable probe also shows the practical mismatch: the exact-base implementation ignores the unused fourth argument, while this head treats it as the new colors output—filling a valid sentinel with 4,389 entries and segfaulting when that stale argument is invalid.

The intended feature itself works in an exact-head build. An external synthetic-plane oracle returns 4,389 aligned points, normals, and CV_32FC4 colors; the legacy two-argument call still returns the same point/normal counts. The complete enabled RGB-D test binary passes 7/7 tests, git diff --check is clean, and the patch merges cleanly with current 5.x. GitHub exposes no checks for this head.

Reviewed with OpenAI Codex assistance; I independently built the exact revisions and verified the public API, source-compatibility, mixed-vtable, runtime color-output, and test results.