Make sure we convert our subprocess input to bytes.

Summary:
Python3 cares that the input is bytes rather than strings.  This is
probably right: even for python2 the filenames could conceivably be
unicode and we should encode them before passing to the subshell.

Test Plan:
Fingers crossed, I don't have a good python3 setup.  But I ran:
    python -c 'print "\n".join(["foo"]).encode("utf-8")'
    python3 -c 'print("\n".join(["foo"]).encode("utf-8"))'
and they returned `'foo'` and `b'foo'` respectively, which seems right
to me.

Reviewers: amos, benkraft

Reviewed By: amos

Subscribers: kevinb

Differential Revision: https://phabricator.khanacademy.org/D63670

hook_lib.py

@@ -29,7 +29,7 @@ def lint_files(files_to_lint):
     # --stdin means we don't need to worry about how many files there are.
     p = subprocess.Popen(['ka-lint', '--stdin', '--blacklist=yes'],
                          stdin=subprocess.PIPE)
-    p.communicate(input='\n'.join(files_to_lint))
+    p.communicate(input='\n'.join(files_to_lint).encode('utf-8'))
     return p.wait()