* Add new contextualized API for hooks and steps
* Make default context configurable
* Run AfterStep hooks even after failed steps, fixes#370
* Update CHANGELOG and README
* Add step result status to After hook, fixes#378
* Elaborate hooks documentation
* Add test to pass state between contextualized steps
* Update README with example of passing state between steps
Best of both worlds, this allows for a single int64 value in the
Options struct, but at the same time we can do `—random` from command
line with no args and have a default value Chosen.
allows for setting randomization with flags, but also modifies the
Options struct to now have two separate values
Overall this relates in a much more understandable UI from a user
perspective, since the normal case for randomization will be a user
just saying “run these out of order please” — and in a rare case they
may wish to additionally supply a seed to try to reproduce an older run
to trigger the same error condition.
while global rand is concurrent safe, since each suite is handling
picking scenario order on it’s own, if running features under
concurrency we cannot guarantee the order Features will execute under,
therefore we have each suite initialize its own randomness Source with
the initial seed, such that their random order selection within a
Feature should always be deterministic when given a specific seed.
This allows for us to propagate the seed value around, and not have to
have two different values in Options when we want to allow
specification of seed. It does introduce some “magic values”, but the
document them in the Options struct docstrings.
Updated lang.feature with the additional cucumber.feature
Updated load.feature with the additional cucumber.feature
Updated events.feature to account for ripple effect of updating lang and load
Updated run.go to indicate that cucumber formatter does not support concurrent.
Updated suite_test.go to handle one new testing step.
