Upgrade to libtorch v1.13.1

[mfl28]

Description of the change

• Upgrades libtorch to 1.13.1 in CMake and Docker setup.
• Updates available CUDA version options (11.6 and 11.7) for libtorch download via CMake.
• The CMake setup now allows to use a range of libtorch versions (>= 1.12.0, <= 1.13.1) when CMAKE_PREFIX_PATH is used.
• Sets linear regression tutorial data dtypes explicitly to float (int-type caused a type-mismatch error with latest libtorch). Also the tutorial now uses the GPU if available - same as other tutorials.
• Updates colab notebook to work with latest libtorch versions. A more recent CUDA version (v11.6 - Colab has incompatible v11.2 by default which causes compilation errors) and more recent g++ are now installed before building the tutorials.

Type Of Change

• New Feature

Related Issues

Closes #105 .

Development & Code Review

• cpplint rules passes locally (run cmake -P cpplint.cmake)
• CI is passing
• Changes have been reviewed by at least one of the maintainers