Timer set fixes

[rkday]

This fixes #216 and adds a UT for it. I also think I spotted a possible race condition (where we peek from a queue, do something with that task, and then pop from the queue, without guarding against the possibility that the head of the queue has changed since we peeked).

I checked that:

• rake spec SPEC=spec/concurrent/executor/timer_set_spec.rb passes
• rake spec SPEC=spec/concurrent/executor/timer_set_spec.rb fails without my commit 894ac23

[coveralls]

Coverage decreased (-7.14%) when pulling af591e0 on rkday:timer_set_fixes into 4e7c5b8 on ruby-concurrency:master.

[coveralls]

Coverage decreased (-7.04%) when pulling 960af94 on rkday:timer_set_fixes into 4e7c5b8 on ruby-concurrency:master.

[jdantonio]

Thank you!

With respect to the failed Travis build, I've never seen that particular test fail on JRuby before. I don't believe it has anything to do with this patch so I've added it to the list of intermittently failing tests listed in Issue #117. We can look into it separate from this pull request.

[jdantonio]

@rkday The failing JRuby tests seems to be a TravisCI issue. Our .travis.yml file specified jruby-19mode as the main JRuby build target. It appears that the fuzzy install that Travis initiates now installs jruby 9.0.0.0-SNAPSHOT for that directive. I've created PR #218 to fix.