need to add some built-in resilience

[sfertman]

Open question right now:

• what happens if atom tries to update and never succeeds?
  • gotta add option for max number of retries before giving up entirely
• what happens if connection to redis is lost? timed out?
  • need to handle those cases gracefully so user has clean api and no surprises
• what happens if something goes wrong in mid transaction? what are the options for rolling back?
  • redis has no ability to roll back multi/exec transactions -- something like this is beyond the scope of this project and probably belongs in refs territory.

[sfertman]

Behaviour will be controlled by an additional :option parameter that may be passed to the constructor. By default will be exactly like clojure atom, for better and for worse.

[sfertman]

take a look at https://github.com/resilience4clj for inspiration

[sfertman]

TODO:

• cas max retries
• cas timeout
  • type of timeout evolution? const, linear, exponential
• swap timeout