Release workflows should spin up temporary AWS instances

[arm4b]

Current Release workflows should spin-up + destroy temporary nodes when it's neeeded, instead of relying on oudated & always running ubuntu-build002 ubuntu-build003 ubuntu-build-itest001 slave nodes.

If provisioning a new node from 0 is not an option, - use at least pre-packed AMIs (start/destroy).

[arm4b]

Related: #283 when there is outdated Mongo 2.4 is installed on those old nodes

[bigmstone]

This also somehow seems far less complex than relying on the build slaves and consul.