Fix batch index task naming convention to accommodate simultaneous tasks

[a2l007]

Description

Modify the batch indexing task ID generation logic from

<index_type_name>_<datasource>_timestamp to
<index_type_name>_<datasource>_<random_id>
This would impact native index tasks as well as hadoop index tasks.

Motivation

With the existing task naming convention, if there are multiple indexing tasks submitted at the same time for a datasource, only one of the tasks gets accepted and the remaining tasks fail with Task[xxx] already exists. I'd like to modify the naming convention to the one mentioned above, so that this issue can be avoided.
@jihoonson Any comments regarding this?

[jihoonson]

@a2l007 thanks for raising this issue! I like it. One question. Do we still need the timestamp at the end of the id? I guess its intention was also to avoid duplicate task IDs.

[a2l007]

@jihoonson Oops sorry, I meant to remove the timestamp from the new convention. I've updated the description accordingly.

[jihoonson]

Ah thanks. It sound good!