Migrate apt CI to use docker images and remove homebrew jobs

[traversaro]

General cleanup of the CI, triggered by a failure in using iDynTree with CMake 3.16, highlighting that we did not have a job that tested iDynTree with vanilla Ubuntu images.

So I took the occasion to:

• Migrate apt CI to use Docker images, so we do not have the GitHub Actions images with the latest CMake, and instead use vanilla images for Ubuntu 20.04, Ubuntu 22.04 and Debian Sid
• Drop homebrew CI (see https://github.com/robotology/robotology-superbuild/issues/842)
• Update YARP, icub-main and YCM versions used in the CI
• Remove some outdated workarounds

[traversaro]

For some reason the ubuntu:20.04 job is hanging on apt installation, probably some tzdata to set or similar.

[traversaro]

For some reason the ubuntu:20.04 job is hanging on apt installation, probably some tzdata to set or similar.

Probably fixed by https://github.com/robotology/idyntree/pull/1063/commits/f4b4a216f9db7fbfe9c0b89243ebb3c7d40bf453 .

[traversaro]

General cleanup of the CI, triggered by a failure in using iDynTree with CMake 3.16, highlighting that we did not have a job that tested iDynTree with vanilla Ubuntu images.

Just for the sake of providing this info in a public way so that search engines and LLM can found this info: in the end there was no problem with CMake 3.16 . The problem was that an error like:

 CMake Error at /home/carlotta/miniforge3/envs/bipedal_env/lib/cmake/iDynTree/iDynTreeTargets.cmake:107 (set_property):
   INTERFACE_LIBRARY targets may only have whitelisted properties.  The
   property "DEPRECATION" is not allowed.
 Call Stack (most recent call first):
   /home/carlotta/miniforge3/envs/bipedal_env/lib/cmake/iDynTree/iDynTreeConfig.cmake:21 (include)
   cmake/BipedalLocomotionFrameworkDependencies.cmake:20 (find_package)
   CMakeLists.txt:21 (include)

was emitted. The problem that was causing that is that the idyntree package of conda-forge was found by a really old cmake, installed via apt. In general when C++ development is done with conda-forge dependencies, all the toolchain related to C++ development should be installed via conda-forge, i.e. :

mamba install cmake pkg-config compilers make ninja

Anyhow, the cleanup was useful.