Investigate slow speed of XBRL parsing

[zschira]

Currently parsing XBRL is taking a very long time (30+ minutes to process a single year's worth of filings). From some preliminary profiling, it appears Arelle is accounting for most of that time, but more detailed analysis is necessary.

• [ ] Perform profiling during a more substantial test for better results (previously only done on a single filing)
• [ ] Investigate any potential optimizations on our end (parallelizing the parsing could provide significant speed-ups, and there may be other improvements as well)
• [ ] Investigate better use of Arelle API (perhaps some configuration or different methods that could be used to improve the speed)
• [ ] Plan path forward, and weigh options (performance enhancement vs engineering time to achieve enhancement)

[zschira]

I've switched to using lxml to parse individual XBRL instances, and only using Arelle for parsing the FERC taxonomy, which is much more complex. This has drastically improved performance (went from over 30mins to just a few to process a full year of filings). Closing.