Consider having consistent data always

[satyamtg]

This has the following changes -

• commented out resource_exists() and its use while downloading as we have consistent data from rsync step
• edited resource_exists() so that scraper doesn't fail on error with requests in a particular URL
• only send filtered URLs to download function, instead of sending all combinations if filtered list is empty (commented that code as can be used in development scenarios)

This fixes #126

[satyamtg]

This is OK and will prevent testing unwanted links but we may want to download those files with a more robust downloader to have extensive retries are we are talking about dozens of thousands of files so the risk for errors (connection mostly) is very high (although mitigated by S3).

The downloads actually occur through curl as seen in https://github.com/openzim/gutenberg/blob/825470482e30bd3ef70e79e587780504c0fc7fa4/gutenbergtozim/utils.py#L101 We might want to have it replaced by save_large_file() from scraperlib as that would eventually be maintained in a much better manner there. Opening an issue for this.