Reduce object creation in Rounding class

[spinscale]

While benchmarking the rounding code due to slower date histogram
aggregations I noticed a few useless object creations due to the
builder/fluent java time API. This reduces creations by properly
creating the objects. Furthermore a few unneeded ZonedDateTime objects
were created in order to create other objects out of them. This was
changed as well.

Running the benchmarks shows a much faster performance for all of the
java time based Rounding classes, which means this was not yet the aggs
date histogram culprit.

[elasticmachine]

Pinging @elastic/es-core-infra

[danielmitterdorfer]

LGTM. Can you please also post the benchmark results here on the PR for future reference?

[spinscale]

benchmark results

Benchmark                                                 Mode  Cnt    Score    Error  Units
RoundingBenchmark.timeIntervalRoundingJava                avgt   30  221,675 ±  1,281  ns/op
RoundingBenchmark.timeIntervalRoundingJoda                avgt   30  377,375 ± 28,692  ns/op
RoundingBenchmark.timeRoundingDateTimeUnitDayOfMonthJava  avgt   30  227,625 ±  5,541  ns/op
RoundingBenchmark.timeRoundingDateTimeUnitDayOfMonthJoda  avgt   30  404,853 ±  3,583  ns/op
RoundingBenchmark.timeRoundingDateTimeUnitJava            avgt   30  134,625 ±  0,859  ns/op
RoundingBenchmark.timeRoundingDateTimeUnitJoda            avgt   30  362,587 ± 23,980  ns/op