src/sage/parallel/map_reduce.py: reduce test dependency on CPU speed

The test we have for start_workers() is doing a bit more than
that. It's starting two threads that each print something, sleep for a
bit, and then print something else. To check the expected output, we
are basically guessing how long it will take these threads to run, and
on a very heavily-loaded machine, we can guess wrong:

  File "src/sage/parallel/map_reduce.py", line 1149, in
  sage.parallel.map_reduce.RESetMapReduce.start_workers
  Failed example:
      sleep(float(1.5))
  Expected:
      Finished: ...
      Finished: ...
  Got:
      Finished: 2

(We were expecting both threads to finish within 1.5s, but only one
did.)

To make this a bit more robust, we eliminate the sleep() and
subsequent print inside of the threads. As a result, we need only make
one guess: how long it will take the threads to start. And for that we
now go overkill and wait 5s.

Author: orlitzky

src/sage/parallel/map_reduce.py

@@ -1136,21 +1136,12 @@ def start_workers(self):
             sage: from sage.parallel.map_reduce import RESetMapReduce
             sage: def children(x):
             ....:     print(f"Starting: {x}", flush=True)
-            ....:     sleep(float(0.5))
-            ....:     print(f"Finished: {x}", flush=True)
             ....:     return []
             sage: S = RESetMapReduce(roots=[1, 2], children=children)
             sage: S.setup_workers(2)
-            sage: S.start_workers(); sleep(float(0.4))
+            sage: S.start_workers(); sleep(float(5))  # long time
             Starting: ...
             Starting: ...
-            sage: [w.is_alive() for w in S._workers]
-            [True, True]
-            sage: sleep(float(1.5))
-            Finished: ...
-            Finished: ...
-            sage: [not w.is_alive() for w in S._workers]
-            [True, True]
 
         Cleanup::