Add S3 support to all input/output files

[junkert]

We need to support AWS-Lambda which will allow Gonymizer to run as a serverless application. To do this we need to be able to store all files in S3 (not just the processed file from the dump command and row_counts file from the dump and load comma). This should be pretty easy to implement since we already have the S3 functions in s3.go

[junkert]

We can use tmp space in lambda jobs. See: https://forums.aws.amazon.com/thread.jspa?threadID=174119

Instead, I found that file storage should be generalized to use either local or s3 storage depending on the path the user supplies on the CLI interface.