(maint) query instances across all regions concurrently

[wkalt]

Query instances across regions concurrently to reduce the execution time
of "puppet resource ec2_instance".

[wkalt]

Timings:

[~/work/puppetlabs-aws] (maint/async-instance-queries) $ time RUBYLIB=./lib puppet resource ec2_instance > /dev/null                                                                                                                          
                                                                                                                                                                                                                                              
real    0m4.366s                                                                                                                                                                                                                              
user    0m1.818s                                                                                                                                                                                                                              
sys     0m0.195s                                                                                                                                                                                                                              
[~/work/puppetlabs-aws] (maint/async-instance-queries) $ git checkout master                                                                                                                                                                  
Switched to branch 'master'                                                                                                                                                                                                                   
Your branch is up-to-date with 'origin/master'.                                                                                                                                                                                               
[~/work/puppetlabs-aws] (master) $ time RUBYLIB=./lib puppet resource ec2_instance > /dev/null                                                                                                                                                
                                                                                                                                                                                                                                              
real    0m16.760s                                                                                                                                                                                                                             
user    0m1.884s                                                                                                                                                                                                                              
sys     0m0.179s

[wkalt]

question: is this the best way to propagate errors from the spawned threads? The downside I see here is that if an error occurs, we are still waiting for all threads to finish before we can raise it.

[wkalt]

@hunner @DavidS can you see if this seems sane when you get a chance?

[wkalt]

travis failures look unrelated to me

[gregohardy]

taking a look. Travis failures don't appear to be related.