compile-time improvements - for_each std::tuple -> std::array

eliminated looping over a tuple and for_each (expensive to instantiate) and to use a straight std::array<std::string_view, ...> kDefaultTags instead. The latter can also be evaluated during compile-time.
added ccache support (can be disabled) which should help for local development that needs to switch between branches.

Comparative timings for the qa_Scheduler unit-tests:

Before (full parallel build):

real    8m43.031s
user    51m13.146s
sys     2m13.891s

ClangBuildAnalyzer output for qa_scheduler before

After:

real    6m52.364s
user    42m5.309s
sys     1m41.661s

ClangBuildAnalyzer output for qa_scheduler after

Visualisation of the remaining traces:

remaining compile-time hot spots:
- BlockRegistry.hpp:25:16: lambda to instantiate each block for a given stream type
- Graph.hpp:323:31: nested lambdas in BlockWrapper::createDynamicPortsLoader()
- fmt::format_to ... rva::variant<...>...: pmt-related printouts
some preliminary findings:
- clang performs 2x better than gcc w.r.t. compile-times but produces similar (the same) asm output
- (especially recursive) lambdas seem to be more compile-time expensive than free-standing functions/methods
- looping over std::array is cheaper as over std::tuple (where applicable).
- initial pre-compiled headers do not seem to have a large positive effect (at least on the chosen unit-test)

[x] simplify BlockRegistry and BlockWrapper to only support minimal API (notably constructors)
[ ] test roll-out <gnuradio-4.0/core.hpp> PCH -> goal: one include dependency for blocklib developer
[ ] use __attribute__((noinline)) to prevent non-performance critical code being inlined/optimised on global scope
[ ] roll out at least three additional pipe-lines for: -fsanitize=address, -fsanitize=undefined, -fsanitize=thread (combine where possible)
[x] centrally suppress plugin system for local/non-runtime builds (e.g. for unit-tests, static/compile-time graphs) ... tbd.
...

fair-acc / gnuradio4