Acquire and ping document event locks in parallel batch

[alechenninger]

Best would be this I think: lightblue-platform/lightblue-rest#244

But alternatively we could use a thread pool and acquire/ping all locks in parallel

[alechenninger]

For an idea of impact, in a non-prod environment, the time spent for each notification was:

• 18ms to find it
• 240ms acquiring, pinging locks
• ?ms doing in memory logic (examing the notification)
• 140ms to mark it processed

By parallelizing lock acquisition and pinging, in a batch of 100 notifications, we could reduce the time per notification by about 54% (to 182ms in this case).

[alechenninger]

Another thought: do we even need locks?

Couldn't we just do:

1. Find notifications
2. Update to status processing if current ... here we will catch if another thread is already processing the notification. We have to do this either way.
3. Any not updated by 2 are dropped since we assume another thread is processing it.