Modify batch index task naming to accomodate simultaneous tasks

[a2l007]

Fixes #8494 .

Description

Modifies the index task name generation logic to as follows:
Before: index_hadoop_datasource_2019-09-27T21:11:29.162Z
After: index_hadoop_datasource_wjkrtyiz_2019-09-27T21:11:29.162Z

---

This PR has:

• been self-reviewed.
• been tested in a test Druid cluster.

---

[gianm]

This isn't enough randomness; RandomIdUtils.getRandomId() offers 32 bits of randomness, but if that's the only thing other than datasource we use for generating a task id, it means that after just a few thousand tasks, odds are pretty good we'll see a collision. It would show up as a task that fails with a duplicate-task-id error. I'd suggest including the current time as well, which would make this a non-issue.

[jihoonson]

@a2l007 thank you for raising this PR! Have you had a chance to check @gianm's comment?

[a2l007]

Sorry for the delay.
@gianm Thanks, that makes sense to me. I'm made the corresponding changes.

[jihoonson]

Sorry for the delay. I will review this PR in a couple of days.

[jihoonson]

I lied. I just read this PR. Looks like it still has the limited randomness issue. How about using UUIDUtils.generateUuid()?

[a2l007]

@jihoonson Thanks for reviewing. Apart from the random string, the timestamp is also part of the task name. Wouldn't that be sufficient to avoid task name collisions?

[jihoonson]

Oh, I missed it. Thanks!