Read fixed chunked size in run_duplicate_streams

This fixes a potential subversion of the timeout parameter
in `process.run_duplicate_streams`.

That is, if the child process writes to much and too fast to its
standard streams, the parent process will take too long (or even
indefinitely) to read and then write the contents to its own
standard streams, before it e.g. checks whether the process should
timeout.

This change adds a size parameter to the relevant read functions
to prevent above.

securesystemslib/process.py

@@ -190,8 +190,11 @@ def _duplicate_streams():
         contents to parent process standard streams, and build up return values
         for outer function.
         """
-        stdout_part = stdout_reader.read()
-        stderr_part = stderr_reader.read()
+        # Read until EOF but at most `io.DEFAULT_BUFFER_SIZE` bytes per call.
+        # Reading and writing in reasonably sized chunks prevents us from
+        # subverting a timeout, due to being busy for too long or indefinitely.
+        stdout_part = stdout_reader.read(io.DEFAULT_BUFFER_SIZE)
+        stderr_part = stderr_reader.read(io.DEFAULT_BUFFER_SIZE)
         sys.stdout.write(stdout_part)
         sys.stderr.write(stderr_part)
         sys.stdout.flush()