If `!search_config.copy_js`, then perform useless search indexing, and search is disabled

Problem

https://github.com/rust-lang/mdBook/blob/3bdcc0a5a6f3c85dd751350774261dbc357b02bd/src/renderer/html_handlebars/search.rs#L42-L54

Here, if !search_config.copy_js, then index is not used, so the above expensive computation is wasted.

More importantly, since the search index is not written, the search feature does not work.
That is, search is only enabled if search_config.enable && search_config.copy_js, instead of the intended behavior of just depending on search_config.enable.

Steps

N/A

Possible Solution(s)

The name copy_js seems to suggest the use of a CDN if set to false, however, looking at https://github.com/rust-lang/mdBook/blob/3bdcc0a5a6f3c85dd751350774261dbc357b02bd/src/theme/index.hbs#L311, the links to a CDN seems to be missing. If so, then the solution would be:

1. Add the CDN links of the search JS to the index.hbs file and enable them when search_config.enable && !search_config.copy_js.
2. Correct create_files to write the index JSON file when !search_config.copy_js.

However, if !search_config.copy_js disabling the search entirely is the intended behavior (which would be weird), then maybe we should just get rid of it and fully rely on search_config.enable.

Notes

Additionally, the search index is unnecessarily duplicated in both searchindex.json and searchindex.js, which are sequentially tentatively queried by searcher.js.

Version

Please see the permalink above pointing to the latest master as of this writing.

Start with src/renderer/html_handlebars/search.rs and inspect the referenced index.hbs and searcher.js paths to understand how copy_js, search_config.enable, and the two index files interact. Determine the intended behavior for copy_js=false, then verify that enabled search loads its assets and index without unnecessary indexing; confirm the generated files and search flow using the repository's existing checks.