Fix call logging by unbuffering stdout/err in call threadpool (#724)

Because we started running method calls in a threadpool, the output was buffered and only flushed/printed at the end (https://stackoverflow.com/questions/18234469/python-multithreaded-print-statements-delayed-until-all-threads-complete-executi). This changes the StreamTee utility we use to print stdout/logging to flush on each write, per https://mail.python.org/pipermail/tutor/2003-November/026645.html

runhouse/resources/provenance.py

@@ -527,6 +527,15 @@ def write(self, message):
  for stream in self.outstreams:
  if message:
  stream.write(message)
+ # We flush here to ensure that the logs are written to the file immediately
+ # see https://github.com/run-house/runhouse/pull/724
+ stream.flush()
+
+ def writelines(self, lines):
+ self.instream.writelines(lines)
+ for stream in self.outstreams:
+ stream.writelines(lines)
+ stream.flush()
 
  def flush(self):
  self.instream.flush()