docs(15.9): add a proximity search page for "word1 word2"~N

[marevol]

Summary

This adds a user page for proximity search ("word1 word2"~N) to the 15.9 tree in all 7 languages (ja, en, de, es, fr, ko, zh-cn). It documents the behaviour added in codelibs/fess#3514, which passes the phrase slop to match_phrase.

Changes

• New page user/search-proximity.rst. It covers:
  • the syntax, title:"..."~N, "..."~ as a normal phrase, truncation of decimal values, and combining with a boost;
  • how N is counted: in analyzer tokens rather than characters or whitespace-separated words, with stopwords still counted and reversed word order needing a larger N;
  • CJK text: the unit depends on the field, so the page advises separating words with spaces and starting with a generous N.
• user/index.rst: adds the page to the toctree after search-fuzzy and mentions proximity search in the intro.
• user/special-char.rst: the ~ row and intro now cover both fuzzy (after a word) and proximity (after a phrase) search. The new page is added to Related Topics.
• user/search-fuzzy.rst: the existing sentence about a number after a phrase now links to the new page, which is also added to Related Topics.

Verification

• I ran a Sphinx build of each language's user/ directory. It produced no new warnings and no unresolved :doc: links. Only the existing "Title underline too short" warnings on CJK sibling pages remain.
• The statements about how N is counted come from measurements on OpenSearch with the default Fess index mapping.