Added support for simulated/embedded gprof output

* This format is seen in simulation of certain embedded
  systems.

* This format does not include a `time` column and contains
  an additional measurement units column underneath (see line
  89 of included example). The units of measurement row
  appears to be ignored without adding any additional ignore
  cases.

* Provided example and test in `tests/prof/embedded.prof`.

gprof2dot.py

@@ -1154,7 +1154,7 @@ def translate(self, mo):
         r'^index\s+%time\s+self\s+descendents\s+called\+self\s+name\s+index\s*$|' +
         r'^\s+called/total\s+children\s*$|' +
         # GNU gprof header
-        r'^index\s+%\s+time\s+self\s+children\s+called\s+name\s*$'
+        r'^index\s+%\s+(time\s+)?self\s+children\s+called\s+name\s*$'
     )
 
     _cg_ignore_re = re.compile(