split SystemCompilerGCC out of SystemCompiler and adopt module_load_environment

[lexming]

This PR fixes three issues with SystemCompiler:

1. fixes using SystemCompiler easyblock for wrapper around system GCC fails if an Intel license is not defined #2815

   It is not possible to have a single class that handles GCC and Intel compilers, because inheritance in python is linear, so as soon as multiple classes are added to SystemCompiler all of them come up in the chain of calls generated by super().
   Solution is to make SystemCompiler agnostic of any compiler family and create child classes per compiler family. This PR adds SystemCompilerGCC for the GCC family. I'm not adding any for icc/ifort as those easyconfigs have been already archived. In the future we could add one for intel-compilers if there is interest.
   This approach is quite simple as the main logic of handling the multiple families remains inside SystemCompiler, avoiding any duplication of code. Child classes only need to set self.compiler_class appropriately.

   This PR is an alternative solution to fix GCC-system without Intel license #3402. The present solution avoids changes in other easyblocks like EB_GCC and is more future-proof as sub-classes handling different compiler families are completely isolated.

2. fixes RPATH wrappers break SystemCompiler easyblock #3558

   RPATH is forced disable in SystemCompilers as it has no role for this type of installation.

3. updates SystemCompiler easyblock for replace make_module_req_guess with module_load_environment in all EasyBlocks #3527

[boegel]

lgtm

[boegel]

Test report by @boegel

Overview of tested easyconfigs (in order)

• SUCCESS GCC-system.eb

Build succeeded for 1 out of 1 (1 easyconfigs in total)
node3506.doduo.os - Linux RHEL 8.8, x86_64, AMD EPYC 7552 48-Core Processor (zen2), Python 3.6.8
See https://gist.github.com/boegel/89b501fe639f2e5a3a7a76603154c133 for a full test report.