Support for S3 "requester-pays" buckets

[menshikh-iv]

Problem:
I trying to open file s3://arxiv/pdf/arXiv_pdf_manifest.xml but get AccessDenied error.

This happens because of arxiv bucket is https://docs.aws.amazon.com/AmazonS3/latest/dev/RequesterPaysBuckets.html.

The idea is simple: you should pay for any queries to this bucket (not bucket owner). For this reason, in tools like awscli or s3cmd authors add --request-payer flag that confrims that you ready to pay money from your account for get access to bucket. For example, see aws/aws-cli#797.

Solution:
Typically, this argument named like {'RequestPayer': 'requester'} or {'x-amz-request-payer': 'requester'} (if we talking about boto3).
I can't pass "object_kwargs" to S3 connector, arguments resource_kwargs or multipart_upload_kwargs aren't suitable for this. For example, RequestPayer='requester' should be passed to any "get-like" method, here https://github.com/RaRe-Technologies/smart_open/blob/a621aebcb34069b0f7850784231c3eaac7e2b964/smart_open/s3.py#L143 and any s3_object.get call.

This code should show content of file after adding "object_kwargs" arguments (instead of exceptions)

from smart_open import open as smart_open

with smart_open("s3://arxiv/pdf/arXiv_pdf_manifest.xml", "r", object_kwargs={'RequestPayer': 'requester'}) as infile:
    print(infile.read())