When `ALTERNATE_WP_CRON` is defined, WordPress runs `wp_cron()` too early [bug report here](https://core.trac.wordpress.org/ticket/24160).

As a result, the `'action_scheduler_run_queue'` hook is triggered before `ActionScheduler_QueueRunner::init()`
has run, because it is hooked to `'init'` with the default priority in `ActionScheduler::init()`.

Currently it will ONLY flush caches if using an external cache and if the filter is not modified, the inverse of what we want.

To ensure that any catchable exceptions that occur while fetching the
action do not block the entire queue (like invalid JSON in `post_content`).

Fixes #26.

To ensure any catchable fatal errors in fetching the action, like that
reported in #26 are handled by ActionScheduler_FatalErrorMonitor (which
only attaches itself on the `'action_scheduler_before_execute'` hook.

Part of #26

Do not include log entries in comment counts

Bugfix: flush cache if not using external cache or if filter is modified

To link to /prospress/ org repo

For real this time

Catch fatal errors when fetching action in queue