vmware datastore selection by scheduler filter/weighter

Allow for Scheduler Filters or Weighters to make the decision as to which datastore the VMware Nova compute driver will use to place a newly provisioned VM.

The problem is that an ESXi host or Cluster of hosts may have access to multiple datastores in which to place their VMDK. When these datastores are all equal, there should be some mechanism to allow the scheduler to decide which datastore it wants to place an image / VMDK. Architecturally, in OpenStack policy level decisions (like which hypervisor to use) occur up at the Scheduler and not down at the driver. To preserve this logical separation of concerns, the driver should report its available datastores to the scheduler and the scheduler should be able to choose the appropriate datastore based on datastore statistics gathered by the driver.

NOTE: this feature should not prevent alternative mechanisms for choosing a datastore (should a user desire to set a parameter for the driver to only use stored named: ^OpenStack.*DataStore$ for example. This should be viewed as an additional feature to add to the stack. If an alternative mechanism is chosen via a setting, the filter's advice should be optionally ignored (with a warning message).