Search: enable precise lang filtering by default

[jtibshirani]

In #60341 we made several improvements to the search-content-based-lang-detection feature. This issue tracks what's needed to complete the feature and enable it by default.

• Symbols service uses ctags to detect the file's language, whereas the search lang filters are based on go-enry languages. So filtering may not work correctly for uncommon languages. We should either (1) switch to storing go-enry langs, or (2) create a really solid conversion function.
• Rockskip does not store languages, so we fall back to the old heuristic path-matching. We should add language detection.
• UX inconsistencies -- fairly rare but ideally we'd tackle these!
  • For some unindexed searches, it's too expensive to return precise language info as part of the results. This leads to inconsistencies between the search query and the filters panel.
  • The blob-view uses an optimization when detecting langs, which can cause inconsistencies between the search query and sidebar.
• Behavior changes -- need to communicate the changes around C/C++ and mitigate effect for customers

/cc @sourcegraph/search-platform

[mmanela]

Related to this https://github.com/sourcegraph/sourcegraph/pull/61301, in this change we need to add special handling to make language matching case-insensitive. This work can be undone / removed once we resolve this item and make precise lang filtering default and remove the older lang filtering.