ci: widen test-distribution macos-test Intel coverage; bump fastmcpp

[Fedr]

Summary

Carves out the Intel-only subset of #6138 so it can land independently:

• test-distribution / macos-test gains two x64 rows alongside the existing macos-15-intel:

  • macos-26-intel (GitHub-hosted, newer macOS)
  • macos-x64-build (self-hosted Mac Pro 2013 on Monterey 12)

  The arm64 row stays at the single macos-latest runner — exactly as in master. No arm64 build host or test runner is touched.

• thirdparty/fastmcpp bumped to c5271bc (clang.hpp + CMakeLists: apply libc++ workaround under AppleClang too fastmcpp#2, already merged into main). That PR widens fastmcpp's _LIBCPP_AVAILABILITY_HAS_INIT_PRIMARY_EXCEPTION 0 workaround so it also fires under AppleClang, both in clang.hpp's #if and at the CMAKE_CXX_COMPILER_ID gate in CMakeLists.txt.

  Without this bump, libMRMcp.dylib (built with AppleClang for the x64 .pkg) references __from_native_exception_pointer — a libc++ helper absent from /usr/lib/libc++.1.dylib on macOS 12. MeshViewer then aborts at load time on Mac Pro 2013 / Monterey:

  dyld: Symbol not found: __ZNSt13exception_ptr31__from_native_exception_pointerEPv
    Referenced from: .../libMRMcp.dylib
    Expected in:     /usr/lib/libc++.1.dylib
  

  The submodule bump is what makes the new macos-x64-build matrix row actually pass.

Why not also include the arm64 changes from #6138

The arm64 side of #6138 expanded the matrix to macos-14 / macos-15 / macos-26 / macos-arm-build-12, and to make macos-14 pass it also had to rebuild Homebrew's cpr formula from source on the arm64 build host (the prebuilt arm64_sequoia cpr bottle ships with LC_BUILD_VERSION minos=15 and references the same __from_native_exception_pointer symbol).

That arm64 rebuild trick only works because our self-hosted arm64 build hosts happen to carry an older Xcode whose libc++ properly suppresses the symbol at the MACOSX_DEPLOYMENT_TARGET=12.7 we pass; on a GitHub-hosted macos-15 arm64 runner (latest Xcode) the same rebuild would reintroduce the symbol. The arm64 path is therefore an undocumented invariant on the build hosts' Xcode version, and worth landing separately once we have a host-independent fix (likely building cpr in MeshLib's own source tree rather than via brew).

This PR sidesteps that whole question — it keeps the arm64 test on macos-latest (a current macOS, where the brew cpr bottle's libc++ symbol does resolve), so no rebuild is needed.

CI scope

Only macOS distribution testing is affected. Non-mac jobs are disabled via labels. full-ci is set so upload_artifacts is true and test-distribution actually runs from build-test-distribute.yml.

Test plan

Verified in run 26288059992:

• All test-distribution / macos-test rows pass:
  • (x64, macos-15-intel)
  • (x64, macos-26-intel) (new)
  • (x64, macos-x64-build) (new — canary for the fastmcpp fix on Monterey)
  • (arm64, macos-latest) (unchanged from master)
• No arm64 build or test configuration has changed vs master.
• All 3 macos-build-test variants succeed; upload-distributions succeeds.

Related

• Subset of ci: widen test-distribution macos-test, rebuild cpr on arm64 only #6138.
• Depends on the already-merged clang.hpp + CMakeLists: apply libc++ workaround under AppleClang too fastmcpp#2.
• Builds on the runner-environment branching in Install Pkg from ci: install pkg to user home on runners without passwordless sudo #6142 (already in master).