search once

[ldwgchen]

Instead of searching the whole document in one go, search and stop at the nearest page that produces a hit.

Fixes #132. UX won't be much different, besides that when the user views a page that hasn't been searched before, they will no longer see highlighted results in this page.

[ldwgchen]

Possible conflict with #450 (see also PR #644) since not all results are obtained at once. However, the proposed implementation does retain results in previously searched pages.

[alerque]

I would consider this a UX regression. I regularly use the search as a way to view the total instances of a key word across book length texts and see the dispersion.

If search results were only incremental this would be much clumsier to achieve. Returning the first result and letting the UI unlock would be nice, but the remaining instances should probably be found and highlighted in the background (and an "N of X" counter in the status bar wouldn't be a bad idea either.

[laomuon]

With the Vim-like command (/), I expect more or less it works like in Vim, i.e. search the whole document and have a counter for it. I agree with @alerque that this could be an UX regression.

[ldwgchen]

I'm personally using zathura-pdf-mupdf. I tried opening a problematic document with mupdf and found the resulting memory usage to be quite similar to what I'd get with zathura-pdf-mupdf. Thus, I believe that the root cause of #132 likely lies upstream. Since search-once is disruptive to the common use cases mentioned by @alerque and @laomuon, it's best to look for an alternative solution and close this PR for now.