Revert of [recipe_modules/step] do not set cwd if it is start_dir.

Revert of [recipe_modules/step] do not set cwd if it is start_dir. (patchset #3 id:10009 of https://codereview.chromium.org/2875543002/ )

Reason for revert:
see comment 20

Original issue's description:
> [recipe_modules/step] do not set cwd if it is start_dir.
> 
> This makes expectations consistent (i.e. steps that are run without cwd
> and steps run in start_dir have the same representation in expectations,
> because they have the same behavior).
> 
> R=dnj@chromium.org
> BUG=
> 
> Review-Url: https://codereview.chromium.org/2875543002
> Committed: https://github.com/luci/recipes-py/commit/c93ed98c3c3e6c9be82e88496e55cb1f3dd13372

TBR=dnj@chromium.org,iannucci@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=

[nodir, 2017-05-17 17:38:40]

The CQ bit was checked by nodir@chromium.org

[nodir, 2017-05-17 17:38:41]

Created Revert of [recipe_modules/step] do not set cwd if it is start_dir.

[commit-bot: I haz the power, 2017-05-17 17:38:43]

CQ is trying da patch.

Follow status at:
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-05-17 17:41:52]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-05-17 17:41:53]

Try jobs failed on following builders:
  Recipes-py Presubmit on luci.infra.try (JOB_FAILED,
https://luci-milo.appspot.com/swarming/task/36304f3f91454710)

[iannucci, 2017-05-17 17:47:09]

This one is going to be hard to revert

[jbudorick, 2017-05-17 17:59:45]

On 2017/05/17 17:47:09, iannucci wrote:
> This one is going to be hard to revert

:(

[iannucci, 2017-05-17 18:03:13]

Something in the recipe looks like it's deleting the current working directory
of the engine. This was previously masked by passing the cwd as a string
duplicate of the original start dir. Going to add some more logging to see if we
can pinpoint when the cwd of the engine gets nuked.

[iannucci, 2017-05-17 18:26:08]

https://codereview.chromium.org/2885293003 should have the same effect, but
should give us more debugging info too.