utils/tdiff.py: Don't parse duplicate targets

When diffing a large target like all, most of the entries in the target
list in query_ninja are duplicated. I haven't looked that closely as to
why this is the case, but uniquing the list brought the size of it down
from 21781 to 76 for me.

Reviewed By: MatzeB

Differential Revision: https://reviews.llvm.org/D156334

Author: lukel97

utils/tdiff.py

@@ -81,7 +81,7 @@ def parse(text):
 def query_ninja(targets, cwd):
     # Query ninja for a node in its build dependency tree.
     proc = subprocess.Popen(
-        ["ninja", "-t", "query"] + targets,
+        ["ninja", "-t", "query", *set(targets)],
         cwd=cwd,
         stdout=subprocess.PIPE,
         universal_newlines=True,