Add support for signing chunked requests

[dblock]

It looks like none of the implementations of the signing interceptor I was able to find supports signing chunked requests with Sigv4. Documentation states that we need a few things:

1. Specify the Transfer-Encoding HTTP header. If you include the Transfer-Encoding header and specify any value other than identity, you must omit the Content-Length header. This is easy and what happens when entity#setChunked(false).
2. For all requests, you must include the x-amz-decoded-content-length header, specifying the size of the object in bytes. This is easy to do, but doesn't actually seem to be necessary for non-chunked requests that come with a Content-Length.
3. Each chunk signature calculation includes the signature of the previous chunk. To begin, you create a seed signature using only the headers. You use the seed signature in the signature calculation of the first chunk. For each subsequent chunk, you create a chunk signature that includes the signature of the previous chunk. I wasn't able to successfully implement this.

Repro in https://github.com/dblock/aws-request-signing-apache-interceptor/tree/repro-20

Long story:

We started from a bug, opensearch-project/OpenSearch#3640, then @jiten1551 reproduced it in https://github.com/jiten1551/Java-Client/tree/master. I forked that to https://github.com/dblock/opensearch-java-client-demo and added code to attempt supporting compressed requests, which seemed to work for a moment. I PRed kumjiten/Java-Client#1, but the next day the same code didn't work, which was also verified by @pgtgrly. I reached out to the AWS SDK team, and followed their recommendation to upgrade to AWS SDK 2.0 with the same results. Then I found this fork. With #31 it now supports Sigv4 signing GET and POST requests, but not chunked requests. We've also added a client setChunkedEnabled setting to explicitly control chunked transfer encoding in opensearch-project/OpenSearch#3864