Workaround a deadlock in the Swift REPL

This was already worked around in rdar://problem/38461035, but it was lost in
a follow-up merge resolution. This reinstates the workaround and adds a test
(from https://bugs.swift.org/browse/SR-7114) to make sure we don't regress this.

apple-llvm-split-commit: ac73e28ffd7b156abbc6a9b3b62db0ba3b4c46f2
apple-llvm-split-dir: lldb/
(cherry picked from commit 510589b)

lldb/source/Plugins/SymbolFile/DWARF/ManualDWARFIndex.cpp

@@ -67,9 +67,11 @@ void ManualDWARFIndex::Index() {
   // done indexing to make sure we don't pull in all DWARF dies, but we need
   // to wait until all units have been indexed in case a DIE in one
   // unit refers to another and the indexes accesses those DIEs.
+  //
+  // This call can deadlock because we are sometimes holding the module lock so
+  // don't do it asynchronously.
   for (size_t i = 0; i < units_to_index.size(); ++i)
-    pool.async(extract_fn, i);
-  pool.wait();
+    extract_fn(i);
 
   // Now create a task runner that can index each DWARF unit in a
   // separate thread so we can index quickly.