perf: do not put closures on the queue

[ForbesLindesay]

This makes the benchmarks slightly faster than version 5.0.0, and way faster than version 6.0.0

[rollingversions]

Change Log for throat (6.0.0 → 6.0.1)

Performance Improvements

• Optimise the queue again

  6.0.0 was significantly slower than 5.0.0. This restores the balance. Each of these benchmarks is run 10x times, in order to make the numbers reasonably stable for comparison.

  5.0.0 benchmark:

  limit=10
  10 promises: 1.118ms
  100 promises: 31.778ms
  1000 promises: 22.806ms
  10000 promises: 59.501ms
  100000 promises: 941.285ms
  1000000 promises: 11.569s
  limit=1000000
  10 promises: 0.177ms
  100 promises: 0.425ms
  1000 promises: 4.904ms
  10000 promises: 84.251ms
  100000 promises: 1.081s
  1000000 promises: 13.656s
  

  6.0.0 benchmark:

  limit=10
  10 promises: 1.122ms
  100 promises: 18.4ms
  1000 promises: 34.742ms
  10000 promises: 86.504ms
  100000 promises: 1.475s
  1000000 promises: 20.298s
  limit=1000000
  10 promises: 0.264ms
  100 promises: 0.424ms
  1000 promises: 5.6ms
  10000 promises: 93.45ms
  100000 promises: 1.466s
  1000000 promises: 20.109s
  

  6.0.1 benchmark:

  limit=10
  10 promises: 1.217ms
  100 promises: 4.433ms
  1000 promises: 49.928ms
  10000 promises: 67.578ms
  100000 promises: 958.481ms
  1000000 promises: 10.542s
  limit=1000000
  10 promises: 0.118ms
  100 promises: 0.424ms
  1000 promises: 4.993ms
  10000 promises: 84.649ms
  100000 promises: 1.163s
  1000000 promises: 13.297s
  

Edit changelog

[coveralls]

Coverage remained the same at 100.0% when pulling f95b825 on perf/match-old-benchmark into f8ced5e on master.