Improve load performance for large datasets

[ryanfb]

After profiling in Chrome, it looks like a large portion of time is spent in calls to .filter, particularly for cite_collection_contains_urn. Perhaps a better approach would be to build up a hash once which we can check directly for every value?

The other large call chain is filter > add_translations > set_cts_text > set_passage > add_urn_li > add_valid_urns > build_cts_ui > success.

Other large hits are in Layout (8s), and Recalculate Style (6s).

[ryanfb]

Chrome load profile on Photios:
Profile-20180201T164512.zip

[ryanfb]

Generating an inverted URN mapping hash makes this go faster, but we still spend a lot of time in page reflow and random jQuery DOM functions. This article seems to suggest that using pure JS and reducing page reflow should vastly improve performance: https://howchoo.com/g/mmu0nguznjg/learn-the-slow-and-fast-way-to-append-elements-to-the-dom

[ryanfb]

Still some (hopefully) minor things to iron out before committing/pushing, but rewriting Photios to pure JS with a single large append into the DOM at the end seems to have significantly improved performance.

Before:

After:

[ryanfb]

Essentially fixed by 6a2a76b, which needs to be merged back and documented.