Perf: Better error report for failed parsing

[marshallward]

Print the line and its tokenization in parse_perf.py if the tokenization fails for any reason. This is most likely to occur if the fourth token is not an integer.

This is possibly a platform-dependent issue, and not something easily replicated on any machine we are using, so it is easier to simply integrate it into the codebase.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 37.12%. Comparing base (6c44c5f) to head (e811635).

❗ Current head e811635 differs from pull request most recent head 0dd57f6. Consider uploading reports for the commit 0dd57f6 to get more accurate results

Additional details and impacted files

@@              Coverage Diff              @@
##           dev/gfdl     #596       +/-   ##
=============================================
+ Coverage          0   37.12%   +37.12%     
=============================================
  Files             0      271      +271     
  Lines             0    80797    +80797     
  Branches          0    15082    +15082     
=============================================
+ Hits              0    29999    +29999     
- Misses            0    45198    +45198     
- Partials          0     5600     +5600     

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

[marshallward]

This did not appear to catch and report the error as hoped, although we did git a bit more information about the problem (ValueError: invalid literal for int() with base 10: 'long)@plt').

Trying a slight modification.

[marshallward]

I was writing the error to stdout which was being piped to a file 🤦. The error is now piped to stderr (or at least that is the intention).