Monitoring parity test: better type comparisons

[ycombinator]

Motivated by failure in https://travis-ci.org/elastic/beats/jobs/517696747:

FAIL: Test that monitoring docs are the same, regardless of how they are shipped.
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/go/src/github.com/elastic/beats/libbeat/tests/system/test_monitoring.py", line 151, in test_compare
    self.assert_same_structure(indirect_beats_stats_doc['beats_stats'], direct_beats_stats_doc['beats_stats'])
  File "/go/src/github.com/elastic/beats/libbeat/tests/system/test_monitoring.py", line 193, in assert_same_structure
    self.assert_same_structure(dict1_val, dict2_val)
  File "/go/src/github.com/elastic/beats/libbeat/tests/system/test_monitoring.py", line 193, in assert_same_structure
    self.assert_same_structure(dict1_val, dict2_val)
  File "/go/src/github.com/elastic/beats/libbeat/tests/system/test_monitoring.py", line 193, in assert_same_structure
    self.assert_same_structure(dict1_val, dict2_val)
  File "/go/src/github.com/elastic/beats/libbeat/tests/system/test_monitoring.py", line 191, in assert_same_structure
    self.assertEqual(type(dict1_val), type(dict2_val))
AssertionError: <type 'float'> != <type 'int'>

In the case of this failure, the type check is a bit too strict. We really only care that the two values are numbers. This PR makes the necessary changes.

Additionally, it also does a bit of cleanup by using isinstance instead of type to determine if a value is a dictionary. Apparently this is the more Pythonic way.

[elasticmachine]

Pinging @elastic/stack-monitoring