Improve performance of PDF viewer after hack introduced

[lhl2617]

See #28

Hack introduced in 0.2.2

[lhl2617]

The current way of point-and-click relies on forcefully rendering the entire PDF into memory, which, as expected, is very memory and computationally intensive. Further, the PDFs are rendered from page [1..n], meaning that the last page will always render last. This is non-ideal..

Optimisations:

• Think of how to incorporate the logic used in Find to do forward PnC
  • This is super high-effort, I hate pure JS, and the PDF.js codebase makes me sad, where types :(
• Backward PnC done for every text layer render (this should be trivial)

If the above optimisations fail:

• Hack again to forcefully render visible views first.

[lhl2617]

I tried today concat-ing the hacked getVisiblePages into a non-hacked one, meaning that the visible ones will render before all of them.

Eh, some mysterious exception--too hacky. When the need requires I'll find a way to do

Optimisations:

• Think of how to incorporate the logic used in Find to do forward PnC

  • This is super high-effort, I hate pure JS, and the PDF.js codebase makes me sad, where types :(

• Backward PnC done for every text layer render (this should be trivial)