VMware: Improve datastore selection logic

VMware: Improve datastore selection logic

The current datastore selection logic is not modularized and difficult
to extend. The retype API implementation has a requirement to specify
hard anti-affinity requirement with the current backing datastore. Some
of the bug fixes also need to specify hard affinity with one or more
datastore types. To support such requirements and to enable future
extensions, this patch introduces a new module which contains datastore
selection logic. The existing code for datastore selection is reused as
much as possible. The dependency on existing datastore selection logic
will be removed in a separate patch.

The current datastore selection iterates over a list of hosts, for each
host, queries the connected valid datastores and tries to select a
suitable datastore. The filtering is based on space, storage profile,
number of connected hosts and space utilization. The space utilization
is used only for breaking ties. If a suitable datastore is found, further
processing of list of hosts is skipped, which could result in uneven space
utilization. To solve this, the new selection logic introduces a requirement
called 'preferred_utilization_threshold' which can be exposed as a driver
config option.

Partial-bug: #1275682
Partial-bug: #1301943
Partial-bug: #1293955

Change-Id: I17e90aa09a303fbb8d4ad90037f440c8c4e7d072