Optimized shortfall analyzer, Autoscaler scale down optimizations

This (non-RC) release introduce the following:

• New optimizing shortfall analyzer for optimal scale up of instances. This can be chosen via scheduling service builder to override the default shortfall analyzer that is aggressive in scaling up more resources than needed.
• Optional callback hooks to Autoscaler now enable you to order the VMs that are idle for scaling down.
• Fix order of releasing of VM leases

Fix order of releasing of VM leases

expire any unknown lease Ids after adding new leases when preparing VMs for scheduling

Assign resource sets in TaskSchedulingService results

• in TaskSchedulingService, assign resource sets before returning result
• optimize debug logs
• new unit tests

Fix cloning of leases for pseudo VMs when using active list

• Fix cloning of leases for pseudo VMs, update totals before use
• Unit test to replicate problem with cloned leases before this fix when using active VMs list

0.13.5-rc.5: Merge pull request #149 from corindwyer/visibility-improvements

add tier resource numbers to no guaranteed capacity message

0.13.5-rc.4: Handle idle/inactive VMs directly in AutoScaler (#145)

The scale down constraint evaluators work with the merged active/inactive lists,
but we need to do extra processing in AutoScaler to enforce min/max idle constraint
for the active server groups.

fix VM cloning for pseudo scheduling

1. check if lease is rejectable before calling reject action.
2. filter out leases from inactive VMs before cloning

Fixes from previous release

• Don't call lease reject action on internal pseudo VMs
• Fix calculation of pseudo VMs to be created in shortfall analyzer
• Make full copy of lease object for cloning pseudo VMs

0.13.5-rc.1

Enable releases to RC repository (#141)

Fix VM cloning

Fixes an issue using new optimizing shortfall analyzer where VMs with scalar resources produced NPE during internal cloning.