Protect from TTL units regressions (see #44)

[thbar]

In this PR:

• one spec to enforce millisecond use on the initial lock creation
• one spec to enforce millisecond use when extending a lock (see extend_life interprets TTL as seconds rather than milliseconds (0.1.8, fixed on master) #44)
• one spec to assert that subsequent calls to extend a lock will reset the TTL rather than add to it

I will work later on another PR (or add to this one if it's not yet merged) to tweak the README accordingly.

[coveralls]

Coverage remained the same at 98.496% when pulling 0ed9683 on thbar:ttl-specs into 7d9c91a on leandromoreira:master.

[leandromoreira]

LGTM

[maltoe]

me too, although I'm generally not a big fan of things like 50.times ... do in the specs. Isn't the "extend resets the TTL" functionality also already covered implicitly by the second spec? :)

[leandromoreira]

ping?

[thbar]

I haven't looked back at the code since July (this was a codebase for a client I'm not working for at the moment), so my memory is a bit fuzzy with regard to the question asked by @maltoe. I'm not a huge fan of 50.times either!

[maltoe]

closing in favor of #61