Draft: fix(search-ranking): add attributes and labels to search ranking

[werererer]

To improve information Retrieval I made some changes to the ranking function:

• Search now considers note label key and values different and tunable in ranking (It didn't before was a bug prob).
• Search now considers the Note-ID
• Made Ranking algorithm more straight forward (likely the first version was similar to the now second)

Note: The search result order will differ due to new rankings. However I compared the ranking with the ranking from the original trilium, and tried to match and improve upon it, by testing it out in practice and hoping that my experience generalizes xD.

The code works in the current stage. But to fine tune it, I will run with it for a bit, tweak it and make it compatible with other wishes of users.

[werererer]

Issues Solved

• Remove Archived Notes from ranking
  [regression] Archived notes are listed in note search/quick search #7085
• PARTIAL SOLVED? (Untested for that particular usecase) Still testing. Caches were introduced to hopefully increase speed.
  Search function slow since 0.98.0 #6808

[werererer]

Associations:

• Improves: feat(search): improve search weights and operators #6536

Dissociations:

• It should not interfer much with. Maybe smaller conflicts to be expected :/ resulting from syntax not semantics
  feat(search): implement FTS5 w/ sqlite for faster and better searching #6839
  OR
  feat(quick_search): just fuzzy match note titles for larger notes, while still matching on exact strings #6810

[werererer]

I am still quite confused tbh, since it is kinda an open question in research how to scale up search reliably: However looking at trilium as a potential folksonomy, the question of how to scale up search is a bit non trivial, but might be solved using equivalence classes instead.

Syntax matching won't scale past a certain note count. Web 2.0 was build using folksonomies. And folksonomies failed using that approach.

One solution might be to actually implement equivalence classes. I am currently drafting ideas here:
#1045

Research pointing out that Equivalence classes might be correct abstraction:

Varese & Castano, 2009, Building Collective Tag Intelligence through Folksonomy Coordination, https://link.springer.com/chapter/10.1007/978-3-642-20344-2_4 [They say that legit equivalence classes might be the solution]

Effective Retrieval of Resources in Folksonomies Using a New Tag Similarity Measure, https://arxiv.org/abs/1207.6033 [They propose clustering, which is just (imo) statistical equivalence classes]

Tag Similarity in Folksonomies, https://inforsid.fr/actes/2013/2013_5_1%20Mousselly%20Sergieh.pdf [Again clustering, and again i believe just statistical equivalence classes]

Conclusion
I will have this open for a bit more, and will work on equivalence classes instead, till a clearer pattern emerges. Maybe there is a generalized solution to scaling up search that no PKM has touched yet.