[service] macos cross-compilation ordering

[ericLemanissier]

What is your problem/feature request?

comes from https://cpplang.slack.com/archives/C41CWV9HA/p1711022464815949

I think there an ordering problem between conan 2 macos-clang and macos-m1-clang runners (cf #23191 (comment))
In case of a package building an executable, the macos ARM (M1) should be built first, because it is used both when targetting the ARM and X86_64 arch. In other words, native compilation has to be scheduled before cross-compilation, which seems to be sequenced backwards currently.
see also
#21622 (comment)
#22378 (comment)

@danimtb @RubenRBS @uilianries

PS: retriggering build never works around this situation, because the order is always wrong.

[ericLemanissier]

The team clearly stated that this won't happen. The following recipes now need contributors to fix them and make sure that the test_package does not tool_requires the tested reference:

libtool/all/test_package/conanfile.py
mbits-lngs/all/test_package/conanfile.py
mfast/all/test_package/conanfile.py
extra-cmake-modules/all/test_package/conanfile.py
grpc/all/test_package/conanfile.py
tree-gen/all/test_package/conanfile.py
flatcc/all/test_package/conanfile.py
btyacc/all/test_package/conanfile.py
protobuf-c/all/test_package/conanfile.py
cfgfile/all/test_package/conanfile.py
libtasn1/all/test_package/conanfile.py
ignition-cmake/all/test_package/conanfile.py
gsoap/all/test_package/conanfile.py
flex/all/test_package/conanfile.py

(list comes from #23594 (comment))

Qt is not worth changing anything, because the recipe does not support cross-compilation, and won't do in a foreseeable future
wayland is not worth changing either, because it supports only linux, so no cross compilation happens on C3I

[AbrilRBS]

Thanks for the recipe traceability on what needs fixing @ericLemanissier - appreciated :)