[libc++] Attempt to fix spurious modules-related failures in the CI

I'm not 100% sure what the issue actually is since I can't reproduce it
locally, however what I explain in the comment is my best attempt to
explain what's going on.

Differential Revision: https://reviews.llvm.org/D92131

libcxx/utils/libcxx/test/features.py

@@ -40,7 +40,15 @@ DEFAULT_FEATURES = [
   Feature(name='has-fobjc-arc',                 when=lambda cfg: hasCompileFlag(cfg, '-xobjective-c++ -fobjc-arc') and
                                                                  sys.platform.lower().strip() == 'darwin'), # TODO: this doesn't handle cross-compiling to Apple platforms.
   Feature(name='objective-c++',                 when=lambda cfg: hasCompileFlag(cfg, '-xobjective-c++ -fobjc-arc')),
-  Feature(name='modules-support',               when=lambda cfg: hasCompileFlag(cfg, '-fmodules')),
+
+  # Note: We use a custom modules cache path to make sure that we don't reuse
+  #       the default one, which can be shared across builds. This is important
+  #       because we define macros in headers files, and a change in these macros
+  #       doesn't seem to invalidate modules cache entries, which means we could
+  #       build against now-invalid cached headers from a previous build.
+  Feature(name='modules-support',
+          when=lambda cfg: hasCompileFlag(cfg, '-fmodules'),
+          actions=lambda cfg: [AddCompileFlag('-fmodules-cache-path=%t/ModuleCache')]),
 
   Feature(name='non-lockfree-atomics',
           when=lambda cfg: sourceBuilds(cfg, """