Add support for reading files from S3

[metasim]

Support for this via the "External File" data source option should be pretty easy and extremely helpful for those of us using AWS EMR for deployment. I attempted to implement this myself, but couldn't figure out how to make sure the hadoop-aws library was included in the local-mode runtime:

https://github.com/s22s/seahorse/commit/909d37852975d88a205d4c3ee98f769e4a3430d9

So I couldn't test the implementation, at least not in local/development mode.

FWIW: The following schemes are supported by org.apache.hadoop:hadoop-aws:

Scheme	Service Provider
s3	org.apache.hadoop.fs.s3.S3FileSystem
s3a	org.apache.hadoop.fs.s3native.NativeS3FileSystem
s3n	org.apache.hadoop.fs.s3a.S3AFileSystem

cc: @mteldridge @mobsy74

[jaroslaw-osmanski]

@metasim some info:

• hadoop 2.7 is shipped with hadoop-aws version supporting s3, s3a and s3n
• current hadoop-aws only supports s3a
• credentials would be passed in core_site.xml (as described here https://www.cloudera.com/documentation/enterprise/5-14-x/topics/cm_s3_clients.html). This means some extra external configuration, but also means that credentials would be protected
• only one s3 account per seahorse is available

[metasim]

@jaroslaw-osmanski Thanks for the additional, helpful info. All that sounds good to us, and we're comfortable with those constraints/requirements.

[metasim]

Proof of concept:

#92