Optimize get_irods_content() for large studies

[mikkonie]

Looking into the current bottlenecks of rendering large studies, I've noticed that, at least in observed cases, the worst is in fact not sheet rendering or loading tables via the Ajax view, but the get_irods_content(). This can take a lion's share of the rendering time when not in edit mode.

I will have to look into this in more detail and see what can be optimized. If decent improvements are feasible with a reasonable amount of work, I'll fit this in the v0.12.1 release. If not, I'll postpone.

Possibilities include e.g. caching this info with sodarcache and retrieving from there in a similar fashion to the planned feature in #1509.

This will most likely involve optimizing the plugin methods called in study and/or app sub-apps.

Related to #956, which may provide one potential solution for improving this.

[mikkonie]

Some findings:

• Building the tables is slow, but is not the main culprit
  • Can be improved by Store sample sheet render tables with sodarcache #1509 but is not exactly critical right now
• For germline studies, the study sub-app get_shortcut_column() is by far the biggest time waster
  • Redundant queries in get_family_sources() which can probably be optimized
  • Optimally, this should come from sodarcache instead of building it every time the sheets are loaded

[mikkonie]

Done, to be merged into dev. The root cause turned out to be calling get_family_sources() once per each row, generating a redundant database query every time. Refactoring this query reduced the execution time in a 5000+ sample study from 115s to 1.2s. D'oh.