fix CacheLocality under gcc on aarch64 with defaults

Summary: By default, gcc on aarch64 uses `std::hardware_destructive_interference_size = 256`. Fix `CacheLocality` to support this value. This could alternatively be tweaked by passing special flags to gcc, but it is simpler to adjust the source file rather than adjust how gcc is invoked when not strictly necessary.

Reviewed By: Orvid

Differential Revision: D68521496

fbshipit-source-id: d5e2cfcaa731961a58622489764e898dbefc2b97

folly/concurrency/CacheLocality.cpp

@@ -500,7 +500,8 @@ class SimpleAllocator {
     allocator->freelist_ = ptr;
   }
 
-  constexpr static size_t kMallocAlign = 128;
+  constexpr static size_t kMallocAlign =
+      std::max(size_t(128), hardware_destructive_interference_size);
   static_assert(
       kMallocAlign % hardware_destructive_interference_size == 0,
       "Large allocations should be cacheline-aligned");