Fix fatal error when job settings not defined (only job runner settings, i.e. jobs run with defaults) in job/queue info.

Stock update job: set internal context before calling processEntities(). (If we updated a new product's stock to 0 because the stock value didn't exist yet, a PHP notice would be emitted.)

Fix: set info['process_*'] only for running processes. This prevents e.g. _drunkins_get_process_config() from returning bogus settings from past processes (e.g. logging) while we're starting a new process.

Standard import jobs: refine isEntityValueEqual() logic. (I hope it does not have adverse effect on existing imports; the '' is equal to NULL is grabbed out of thin air. I don't want either to be equal to 0 though.) Also, get default values from database schema and do not set 'changed' flag if import value is NULL and entity value is the default.

Implement drush command to start a process and run a full process; change lots of logs strings so that they don't output HTML in drush console output.

Fix some more bugs in the UI & logging, and implement one TODO about setting different logging settings during start vs. rest of the process (which was already possible).

Implement new 'method' field in status info / table; renumber 'running' and 'run_info' items. Document an extension of run_info for queue-like processes but do not implement it yet.

Various unrelated bugfixes / changes, while working on the previous 3 commits. Also, move _drunkins_s/get_process_info() into run.inc (unchanged), because we do a module_load_include at the top of every function anyway (because we keep forgetting that _drunkins_log() is in run.inc too). And delete some deprecated config.

Replace 'override_fetch' setting in drunkins_run_start() by a separate argument because I'm tired of deciding whether it's "not a job setting" or "not a runner setting". (And having it as a job setting means namespace issues.)

Make the standard product import job capable of triggering a stock update (using the standard stock update job) whenever a new product is imported.

* Make drunkins_enqueue_items() support a new 'can_override_fetch=pass_through_start(_by_default)' capability / proper logging after queueing the items  / enqueue_by_fetcher. The capabilities are not fit for use yet, though (see lots of @todos about locking).
* Squash bugs in drunkins_enqueue_items().
* Change 'skip_last_marker' runner setting to 'add_last_marker', which is clearer - now that we decided to make it have different default in different functions.

Make a plan for tackling complications around doing certain calls on hook_cron_queue_info() vs hook_cron(). Move the commented-out buggy _drunkins_queue_process_check_finish() call from the former to the latter, so we can uncomment it - and officially document the 'process_items_after_rerun_start' setting as "not working / processes won't start" for now.

Some fixes to last year(!)s importers rewrites; some @todo/warnings in the runner. After commenting out some dangerous code, it is now possible that processes do not properly finish anymore...

Change _drunkins_get/set_process_config, remove the 'running_or_starting_process' argument value (I truly don't know what the exception for force_cron/progressive settings was supposed to be; force_cron should certainly be stored.) Change the way 'override_fetch' (which isn't really a setting) is handled. 
Also use the calls more consistently.

Fix return value from drunkins_cron_queue_info() to work with elysia_cron 2.1. In fact, only return queue info for processes that should be running from now on; this is more understandable and makes us need only one large if() statement.

INCOMPATIBLE CHANGE: redefine job interface(s) to include @throws and make the runner code handle them. For now, back out settingsForm() from the official interface specification.

Fix 'enqueue_by_fetcher' handling (until now, jobs were always terminated upon start!) plus 2 more bugs introduced somewhere in the past months.

- fix drunkins_get_job() so it can handle 'process_runner_settings' being empty (bug introduced recently)
- fix some other code to properly handle finishing off a failed batch process.

Implement log level and properly implement log_severity_no_items. Make clearer that 'logging' is a runner setting as far as our code is concerned (though we still use it as a 'direct job setting' in cases we don't use a runner.)

Now that run settings are used more consistently, create separate functions to get/set them and don't pass them through function parameters anymore. Change first parameter to logging function. Also, move drunkins_get_job() into .inc file because it's an internal function.

Hopefully fix a bug in HEAD^^ drunkins_start_job() where $info['process_runner_settings'] was not set.

Half baked support for 'enqueue_by_fetcher' run setting, with lots of @todos. Works on cron (if you hack your fetcher/job to do DrupalQueue::get($this->settings['job_id'])->createItem($item) by itself), not for batch processes.

Rename 'run options' to 'runner settings' and use them a bit more consistently: move 'class' and 'fetcher_class' into them. (Meaning you can theoretically call drunkins_run_start() to run a queue job without defining config in an info hook.) hook_queue_info() definitions now have a 'runner_settings' section but for the rest, code behavior should not have changed.