Add support for warmups and iterations for each query

[jihoonson]

This PR proposes to add support for two new optional arguments, warmup_iterations and iterations.

• warmup_iterations specifies the number of warmup iterations for each query to run before its performance is measured. The warmup timings are not reported. This is set to 0 by default.
• iterations specifies the number of iterations for each query to run and measure its performance. This is set to 1 by default.

An example snippet of the report with warmup_iterations=3 and iterations=5:

...
local-1741649529813,CreateTempView catalog_sales,2245
local-1741649529813,CreateTempView store_sales,1697
local-1741649529813,query1,874
local-1741649529813,query1,952
local-1741649529813,query1,763
local-1741649529813,query1,860
local-1741649529813,query1,752
local-1741649529813,Power Start Time,1741649555
local-1741649529813,Power End Time,1741649565
local-1741649529813,Power Test Time,10000
local-1741649529813,Total Time,38532
...

But the query1 ran 8 times as intended (3 warmups, 5 actual runs).

These new arguments are optional, and thus they do not change existing behaviors when they are not set. When they are set, the power run script runs each query warmup_iterations times first, and then iterations times to measure timings. Only the timings after the warmup are reported.

Unlike the existing runs argument that specifies the number of power runs to run, these new arguments specify per-query iterations. In Spark terms, the runs argument can submit multiple Spark applications where each application performs one power run, while the new arguments specifies the number of iterations for each query to run in the same Spark application. This could be useful especially when you want to keep some states between multiple query runs in Spark, such as file cache, to get more consistent results.

[abellina]

LGTM