Decouple authenticator and socket creation

[ankona]

1. ZMQ authenticators appear to have clashing inproc addresses when using the zmq.Context.instance() factory method. Replaced as needed.
2. Updated underlying Dragon library version, which included a breaking changing causing the swap from TemplateProcess to ProcessTemplate
3. Fixed incomplete permission set on curve key files

[codecov]

Codecov Report

Attention: Patch coverage is 23.71134% with 74 lines in your changes are missing coverage. Please review.

❗ No coverage uploaded for pull request base (dragon_launcher@547e20a). Click here to learn what that means.

❗ Current head ba0a0e5 differs from pull request most recent head d8df7a6. Consider uploading reports for the commit d8df7a6 to get more accurate results

Additional details and impacted files

@@                Coverage Diff                 @@
##             dragon_launcher     #542   +/-   ##
==================================================
  Coverage                   ?   63.66%           
==================================================
  Files                      ?       75           
  Lines                      ?     5403           
  Branches                   ?        0           
==================================================
  Hits                       ?     3440           
  Misses                     ?     1963           
  Partials                   ?        0           

Files	Coverage Δ
smartsim/_core/control/job.py	90.00% <100.00%> (ø)
smartsim/_core/launcher/launcher.py	100.00% <100.00%> (ø)
smartsim/_core/utils/telemetry/manifest.py	93.02% <100.00%> (ø)
smartsim/_core/launcher/dragon/dragonBackend.py	0.00% <0.00%> (ø)
smartsim/_core/launcher/dragon/dragonSockets.py	32.50% <16.66%> (ø)
smartsim/_core/utils/telemetry/telemetry.py	77.51% <0.00%> (ø)
smartsim/_core/utils/security.py	41.73% <28.00%> (ø)
smartsim/_core/launcher/dragon/dragonLauncher.py	25.56% <9.09%> (ø)

[MattToast]

Throwing some initial feedback your way! In general I think the authenticator changes look good, but I noticed that this PR was also making some changes along side the launchers and the telemetry monitor that didn't look strictly related. Is there anyway we can break them out into their own PRs in case we need to roll one back?

[MattToast]

I'm not sure how I feel about shallow copying runs and exp_dir out of manifest_dict but then passing manifest_dict around the everywhere alongside the copied data. Can we just pass around the manifest_dict instead??

[ankona]

I wanted to give the deserialization the object it was working on but also let it see higher level context info when it needed to do something wacky, like change behavior if it's a dragon launcher. I didn't like sending is_dragon all the way down, nor do I like the idea of having to add a launcher: str param, then another later, etc...

I think I like switching to passing only the experiment. That achieves my goal but avoids passing the entire manifest around. What do you think of that?

[MattToast]

That makes sense to me, and resolves my "multiple parameters pointing to same data" complaint! Totally in favor!!

[MattToast]

LGTM pending tests!! Thanks for cleaning this up!!

[al-rigazzi]

L
G
T
M