Enhance regression test of dynamic buffer calculation (#3694)

- What is the motivation for this PR?
Enhance regression test of dynamic buffer calculation

- How did you do it?
* Adjust test cases for 8-lane ports. On some switches, ports with 8 lanes need special handling. 
  Port with/without 8 lanes will have different buffer profiles when they share the same speed.
  Number of lanes needs to be taken into account when generating the expected profiles.
  Make port_to_test a module-level fixture and introduce new global NUMBER_OF_LANES.
  This is to hide the details regarding the number of lanes into the method to generate the expected profile.
* Adjust buffer pool calculation according to private headroom
* Adjust expected error messages of test_exceeding_headroom
* Move some error messages to ignore list as they don't always happen in time during the test.
* Only error messages reported by buffermgrd are in the expected list

- How did you verify/test it?
Run regression test.

Signed-off-by: Stephen Sun <stephens@nvidia.com>

tests/qos/files/dynamic_buffer_param.json

@@ -30,7 +30,8 @@
 	    "default": "19456"
 	},
 	"shared-headroom-pool": {
-	    "size": "1024000"
+	    "size": "1024000",
+	    "private_pg_headroom": "10240"
 	}
     }
 }