[Bugfix][VTA] Fix FSIM compile error on macOS.

VTA FSIM could not be built on macOS, for it leverages malloc.h and
memalign, yet both have been deprecated and are not provided by
macOS. This issue was captured in apache#13173.

This commit stops including malloc.h in VTA Runtime as stdlib.h has
provided functions we need.

This commit uses posix_memalign instead of memalign. It is a portable standard function.

vta/runtime/runtime.cc

@@ -27,7 +27,6 @@
 #include "runtime.h"
 
 #include <dmlc/logging.h>
-#include <malloc.h>
 #include <stdlib.h>
 #include <tvm/runtime/c_runtime_api.h>
 #include <vta/driver.h>
@@ -76,7 +75,12 @@ class AlignmentAllocator : public std::allocator<T> {
 
   inline const_pointer address(const_reference r) const { return &r; }
 
-  inline pointer allocate(size_type n) { return (pointer)memalign(N, n * sizeof(value_type)); }
+  inline pointer allocate(size_type n) {
+    pointer mem = nullptr;
+    const int err = posix_memalign((void **) &mem, N, n * sizeof(value_type));
+    ICHECK_EQ(err, 0) << "InternalError: failed to allocate aligned memory. ";
+    return mem;
+  }
 
   inline void deallocate(pointer p, size_type) { free(p); }
 
@@ -528,7 +532,9 @@ class UopQueue : public BaseQueue<VTAUop> {
       total_size += ksize;
     }
 
-    char* lbuf = (char*)memalign(ALLOC_ALIGNMENT, total_size);
+    char* lbuf = nullptr;
+    const int err = posix_memalign((void **) &lbuf, ALLOC_ALIGNMENT, total_size);
+    ICHECK_EQ(err, 0) << "InternalError: failed to allocate aligned memory for load buffer. ";
     uint32_t offset = 0;
     for (uint32_t i = 0; i < cache_.size(); ++i) {
       uint32_t ksize = cache_[i]->size() * kElemBytes;