Add looped mode to bulk operation

[gbanasiak]

Adds looped mode to bulk operation. That is useful in benchmarks where we want Rally to ingest for a specific amount of time even if corpus is not big enough to sustain indexing at a specific rate for this time.

This mode is meant to be used together with time-period or iterations at the task level. Otherwise Rally will never complete the task.

Example use:

{
  "name": "some-bulk-task",
  "operation": {
    "operation-type": "bulk",
    "bulk-size": 5000,
    "looped": true
  },
  "time-period": 300,
  "target-throughput": 2
}

[gbanasiak]

Re-running IT test failure of which seems unrelated.

[inqueue]

LGTM

[inqueue]

With looped mode, we will have the possibility to add an override for bulk and batch sizes in steps if it is something we can use in the future.

Edit: I see bulk size would be relatively easy to adjust in the task definition itself, and not necessarily here directly.

[gbanasiak]

The idea is looped operation should be limited at the task level, and you can have as many tasks with different batch sizes, etc. as you want. Looped mode basically means "keep on generating bulk requests until some external criteria stop the task".