Various ENH + RF +BF to make aws_ec2 and ec2_condor (more) usable

[yarikoptic]

Individual commits have more information. Changes are quite sloppy but the goal was to make things work:

• the whole dance started with trying to establish a new ec2_condor and it asking me for which key_name I want as many times (in parallel) as the size ;) now it would not ask at all!:
• asking for key_name and breeding keys is not nice/good -- now key matches the name by default and we "mint" key_filename to match
  • now if we are to ask for a key_name, we would offer to choose from already present locally keys
• ec2_condor was not really usable without explicit specification of base dir to reside somewhere under ~/nfs-mount. Now should function "by default" since ~/.reproman would be symlinked from that location
• unresolved stumbling point is https://github.com/datalad/datalad/issues/5382 forbidding fetching outputs (and probably upload of inputs if not available "publicly"). For now it requires adding to ~/.ssh/config entry like

  Host 54.166.157.42
  IdentityFile /home/yoh/.local/share/reproman/ec2_keys/my-nitrc.pem 

Closes #567

ATM running a "realistic" test case which entailed (work towards fulfilling #548)

• reproman create yoh-aws-hpc2 -t aws-condor -b size=2 -b key_name=my-nitrc -b instance_type=t2.medium to establish 2 node cluster
• reproman run -r yoh-aws-hpc2 --sub condor --orc datalad-pair --jp "container=containers/bids-mriqc" --bp subj=02,13 --follow --input 'sourcedata/sub-{p[subj]}' --output . '{inputs}' . participant group -w workdir --participant_label '{p[subj]}' to mimic the example of https://github.com/ReproNim/containers#a-typical-workflow . Tomorrow will see if/where/how it fails

[yarikoptic]

crap -- it took travis a night to run the tests, thought to migrate but somehow the repronim organization does not even give me an option to grant or request, like it was may be denied at some point?

Separate issue discovered (unrelated): https://github.com/ReproNim/reproman/issues/568

I have pushed e5c3b80c7f03b114548113ac0e02694cf546c8ff so we install released datalad, lets see also if above (and the log flood https://github.com/datalad/datalad/issues/5386 ) improves.

But overall I consider this in usable state without immediate side-effects