Add additional samples and ensure that git does not modify samples

[nhairs]

This PR:

• adds the sample from #429 to prevent regressions.
• adds a copy of dmarc_ruf_report_linkedin.eml with line endings converted to CRLF to ensure there is at least one forensic sample with the line endings.
• adds a .gitattributes file that ensures that the samples/ directory is treated as binary data. Without this git would treat this data as text data and automatically convert between /r/n (CRLF / DOS) line endings and /n (UNIX) line endings when files are checked out. This is important for ensuring that our parsing logic can be tested on both types of line endings.

Test Plan

Mimecast Regression

1. Checkout 1de4a94c3787d2470e4dacd8e0de1b4027bc0f21 (parent of commit that fixed mimecast issue)
2. Run tests and check that it fails with parsedmarc.InvalidAggregateReport: Invalid archive file: Not a gzipped file (b'\r\n')
3. checkout HEAD and run tests again, this time with no errors.

Forensic sample with CRLF

1. Run file samples/forensic/dmarc_ruf_report_linkedin.crlf.eml
2. Check that the output says that it has "CRLF line terminators"

Aggregate sample with CRLF

1. Run file samples/aggregate/mimecast-weird-gzip.eml
2. Check that the output says that it has "CRLF line terminators"

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (100f12e) 58.42% compared to head (a93fa08) 58.42%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #456 +/- ## ======================================= Coverage 58.42% 58.42% ======================================= Files 11 11 Lines 1347 1347 ======================================= Hits 787 787 Misses 560 560 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[seanthegeek]

Thanks!