age: scrypt fix work factor calculation with bad SystemTime precision

[Rileran]

This PR fixes issue #418

Uses default scrypt work factor if the duration of one scrypt encryption is below SystemTime precision.

[str4d]

Rather than falling back to the default (which in this case could be significantly faster than 1 second on a platform that is this fast), what I'll do is re-run the test with larger log_n values until the measured duration is non-zero. It stands to reason that this will impose the minimum cost on encryption, as any run that measures below SystemTime precision can be considered negligible.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (488212d) 40.81% compared to head (c2b91c9) 41.73%.
Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #419      +/-   ##
==========================================
+ Coverage   40.81%   41.73%   +0.92%     
==========================================
  Files          34       34              
  Lines        3249     3225      -24     
==========================================
+ Hits         1326     1346      +20     
+ Misses       1923     1879      -44     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[str4d]

Closing in favour of #424.