Support reconnecting on the specified error

[yoheimuta]

Adds a support for a known issue in ElastiCache failover. We must wrap all sync write calls and async wait calls without the library support. It costs a speed penalty by boilerplates and missings by mistakes.

I'm posting this to get more feedbacks on the output because this is large and the detail is not issued yet.

See

• Issues in ElastiCache Redis
  • http://qiita.com/dany1468/items/8946cd5e4c853b48bffd -- in japanese
  • https://github.com/luin/ioredis/issues/144
• An interface design example
  • https://github.com/luin/ioredis#reconnect-on-error

I expect that this pr tests are passed after #56 is merged.

[yoheimuta]

ping. (because time has passed since 9/11)

[shogo82148]

Sorry for my late response. LGTM 👍 I will fix some other bugs, and then release within a few days.

[yoheimuta]

Thanks!