Improve CallDataCopyOperation worst case

[ahamlat]

PR description

Improve a worst case scenario performance spotted by EEST memory tests on CallDataCopy opcode.
Surprisingly, the benchmarks showed that CallDataCopy performs worst when there's no actual data to copy, and this is related to the overhead of decoding the 3 values popped from the stack.

This PR implements an improvement for this case (data size = 0), to not decode the source offset if the number of bytes to copy is 0, and return directly the result.

Before this PR

Benchmark                                              (dataSize)  (fixedSrcDst)  (nonZeroData)  Mode  Cnt       Score      Error  Units
CallDataCopyOperationBenchmark.executeOperation                 0          false          false  avgt   15      97.077 ±    1.487  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0          false           true  avgt   15      96.577 ±    1.966  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0           true          false  avgt   15      88.163 ±    1.086  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0           true           true  avgt   15      88.822 ±    1.711  ns/op
CallDataCopyOperationBenchmark.executeOperation               100          false          false  avgt   15     182.074 ±    8.460  ns/op
CallDataCopyOperationBenchmark.executeOperation               100          false           true  avgt   15     183.084 ±    4.897  ns/op
CallDataCopyOperationBenchmark.executeOperation               100           true          false  avgt   15     119.650 ±    9.987  ns/op
CallDataCopyOperationBenchmark.executeOperation               100           true           true  avgt   15     113.201 ±    0.695  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240          false          false  avgt   15    1117.707 ±   15.820  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240          false           true  avgt   15    1117.801 ±    7.763  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240           true          false  avgt   15     312.852 ±   39.833  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240           true           true  avgt   15     304.982 ±   35.716  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576          false          false  avgt   15  133887.226 ± 2171.843  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576          false           true  avgt   15  134614.670 ± 1066.297  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576           true          false  avgt   15   24759.562 ± 4016.988  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576           true           true  avgt   15   23758.696 ± 2273.540  ns/op

With this PR

Benchmark                                              (dataSize)  (fixedSrcDst)  (nonZeroData)  Mode  Cnt       Score      Error  Units
CallDataCopyOperationBenchmark.executeOperation                 0          false          false  avgt   15      71.941 ±    1.762  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0          false           true  avgt   15      78.568 ±   12.322  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0           true          false  avgt   15      66.725 ±    4.050  ns/op
CallDataCopyOperationBenchmark.executeOperation                 0           true           true  avgt   15      64.491 ±    2.025  ns/op
CallDataCopyOperationBenchmark.executeOperation               100          false          false  avgt   15     184.060 ±    4.993  ns/op
CallDataCopyOperationBenchmark.executeOperation               100          false           true  avgt   15     180.937 ±    9.094  ns/op
CallDataCopyOperationBenchmark.executeOperation               100           true          false  avgt   15     112.054 ±    2.260  ns/op
CallDataCopyOperationBenchmark.executeOperation               100           true           true  avgt   15     114.233 ±    1.935  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240          false          false  avgt   15    1186.422 ±  187.566  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240          false           true  avgt   15    1124.052 ±   43.443  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240           true          false  avgt   15     304.100 ±   24.339  ns/op
CallDataCopyOperationBenchmark.executeOperation             10240           true           true  avgt   15     279.107 ±    2.298  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576          false          false  avgt   15  135896.704 ± 1630.046  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576          false           true  avgt   15  133703.610 ± 1657.089  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576           true          false  avgt   15   22804.807 ±  567.671  ns/op
CallDataCopyOperationBenchmark.executeOperation           1048576           true           true  avgt   15   22540.610 ±  571.624  ns/op

There is between 18 and 27 % improvement on zero data use case.

Configuration	Before (ns/op)	After (ns/op)	Improvement	Percentage
fixedSrcDst=false, nonZeroData=false	97.077	71.941	25.136 ns	25.9%
fixedSrcDst=false, nonZeroData=true	96.577	78.568	18.009 ns	18.6%
fixedSrcDst=true, nonZeroData=false	88.163	66.725	21.438 ns	24.3%
fixedSrcDst=true, nonZeroData=true	88.822	64.491	24.331 ns	27.4%

Fixed Issue(s)

Thanks for sending a pull request! Have you done the following?

• Checked out our contribution guidelines?
• Considered documentation and added the doc-change-required label to this PR if updates are required.
• Considered the changelog and included an update if required.
• For database changes (e.g. KeyValueSegmentIdentifier) considered compatibility and performed forwards and backwards compatibility tests

Locally, you can run these tests to catch failures early:

• spotless: ./gradlew spotlessApply
• unit tests: ./gradlew build
• acceptance tests: ./gradlew acceptanceTest
• integration tests: ./gradlew integrationTest
• reference tests: ./gradlew ethereum:referenceTests:referenceTests
• hive tests: Engine or other RPCs modified?