Add SourceFingerprint based on path, check_name, and source

[dblandin]

Currently, the default fingerprint for an issue is computed by creating an MD5 digest based on the path of the file which introduced the issue and the check name for the issue emitted.

The result of this fingerprint for many engines is that issues emitted with the same check name for the same file are given the same fingerprint. On codeclimate.com, we only show issues with unique fingerprints, so even if 100 issues were emitted for a file, if all issues share the same check name and thus the same fingerprint, we only show one issue.

Additionally, a fixed or introduced issue is often ignored if there are current issues with the same fingerprint already on the default branch.

This PR introduces new fingerprinting behavior to address these issues, first added to a couple engines, that incorporates relevant source code into the fingerprint. This implementation allows us to enable the new fingerprinting behavior with an environment variable so we can roll it out gradually. Eventually, this will become the new default fingerprinting method for engines.

@codeclimate/review 🔎

[pbrisbin]

I think this is telling you these objects should live under the Analyzer namespace. Is there a reason you went with CLI?

[dblandin]

Ahh, that was an oversight. I'll move it into Analyzer.

[dblandin]

@wfleming I pushed up a commit with support for location positions either in the line/column or offset form. Let me know what you think.

[dblandin]

@codeclimate/review This is ready for another look! I addressed all of the comments brought up in the last round.

[pbrisbin]

FWIW: I've started to come around to:

offset +=
  if line == index
    column
  else
    source_line.length
  end

Which appeases rubocop without misaligning the if/else/end depending on the size of the LHS of the assignment.

[dblandin]

Looks a little strange, but seems better to me! 👍

[pbrisbin]

Some minor comments. Looks good enough to RC and move onto a builder branch IMO.