let F14Table::size explicitly be size_t

Summary:
Closes: facebook/folly#2551.

The `F14Table::iterator` type packs the size in with other table metadata into `uint64_t`. But it is possible on some platforms for `size_t` to be smaller than `uint64_t`, making some uses of the result of `size()` into implicit narrowing casts.

The size of the table cannot be larger than is representable by `size_t` but due to the bit-packing and shifting we still need the `uint64_t` representation. So let `F14Table::size()` explicitly cast.

Reviewed By: r-barnes

Differential Revision: D93896663

fbshipit-source-id: 39e55b8d21056fcd43b0fdba942603be9953634b

Author: yfeldblum

third-party/folly/src/folly/container/detail/F14Table.h

@@ -1579,7 +1579,9 @@ class F14Table : public Policy {
 
   bool empty() const noexcept { return size() == 0; }
 
-  auto size() const noexcept { return sizeAndChunkShiftAndPackedBegin_.size(); }
+  std::size_t size() const noexcept {
+    return to_narrow(sizeAndChunkShiftAndPackedBegin_.size());
+  }
 
   std::size_t max_size() const noexcept {
     auto& a = this->alloc();