[v2] Benchmark harness improvements & add more service benchmarks

[aemous]

Issue #, if available:
N/A

Description of changes:

• Adds support for specifying file literals in the benchmark definitions JSON.
• Updated README with documentation on specifying file literals.
• Added dynamodb benchmarks for get-item, batch-get-item, put-item commands, and a benchmark for the customization ddb put.
• Aded ec2, ecs, cloudwatch logs, and cloudwatch benchmarks.
• Added STS benchmarks for AssumeRole, GetCallerIdentity, AssumeRoleWithWebIdentity.
• Removed S3 benchmarks using the CRT client because they are currently non-functional with a non-trivial fix. Will possibly revisit getting them working later.
• Fix bug in benchmark summarizer with computing averages.
• Add support to the benchmark harness for debugging the child process.
• Redirect child process output to files so that the only output of run-benchmarks script on success is the benchmark results.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[ashovlin]

How much does the file:// usage impact this output as opposed to specifying the key inline? I see we use it in examples, but curious if it's worth separating DDB's performance from that of file://? But if it's relatively small compared to the full operation, okay with this.

[aemous]

One argument for file usage is so that we can have performance testing coverage of that feature of the CLI.

I can see the value of keeping it out if possible. We can also add it as a dimension, "KeyLoadedFromFile": "True".