autoscaler does not consider whether pod can schedule to node causing pending status

[typhoonzero]

I'm running case1 in #399

Non-terminated Pods:		(6 in total)
  Namespace			Name				CPU Requests	CPU Limits	Memory Requests	Memory Limits
  ---------			----				------------	----------	---------------	-------------
  kube-system			calico-node-qzn36		250m (2%)	0 (0%)		0 (0%)		0 (0%)
  wuyi05-baidu-com		mnist14-trainer-k22dp		2 (16%)		3 (25%)		8Gi (8%)	8Gi (8%)
  wuyi05-baidu-com		mnist15-trainer-z4s0r		2 (16%)		3 (25%)		8Gi (8%)	8Gi (8%)
  wuyi05-baidu-com		mnist17-trainer-wqrpz		2 (16%)		3 (25%)		8Gi (8%)	8Gi (8%)
  wuyi05-baidu-com		mnist5-pserver-fk552		2 (16%)		3 (25%)		5Gi (5%)	5Gi (5%)
  wuyi05-baidu-com		mnist9-pserver-vb1nz		2 (16%)		3 (25%)		5Gi (5%)	5Gi (5%)
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests	CPU Limits	Memory Requests	Memory Limits
  ------------	----------	---------------	-------------
  10250m (85%)	15 (125%)	34Gi (36%)	34Gi (36%)

node CPU Request is 10250m and pod request 2 cores (2000m), autoscaler is still scaling up in this case.

[Yancey0623]

Thanks, I think autoscaler should compare with the idel resource and the job requests before scale up.
Will fix this bug.