fix(fts): honor custom ignore_pattern (and tokenizer) when normalizing queries

[adsharma]

Fixes LadybugDB/ladybug#910

Problem

CREATE_FTS_INDEX accepts an ignore_pattern option, but the rewritten CALL _CREATE_FTS_INDEX(...) only forwarded stemmer and stopWords. The index itself was built with the custom pattern (via the TOKENIZE macro), but the catalog entry that QUERY_FTS_INDEX reads kept the default pattern — so any token the custom pattern was meant to preserve (digits, hyphens, …) could never be matched by an exact query:

CALL CREATE_FTS_INDEX('T', 'idx', ['content'], stemmer := 'none', ignore_pattern := '[^[:alnum:]-]+');
CALL QUERY_FTS_INDEX('T', 'idx', 'A4') RETURN *;    -- [] before, now matches
CALL QUERY_FTS_INDEX('T', 'idx', 't-roc') RETURN *; -- [] before, now matches

tokenizer and jieba_dict_dir were dropped the same way, which additionally made jieba indexes tokenize queries with the simple tokenizer.

Changes

• create_fts_index.cpp: forward ignore_pattern, tokenizer and jieba_dict_dir to _CREATE_FTS_INDEX so the persisted FTSConfig matches the index that was actually built.
• fts_utils.cpp (normalizeQuery): new optional wildcard protection — the query is normalized in segments between */? characters so wildcards survive even when the pattern would match them (e.g. a negated class like [^[:alnum:]-]+). The previous trick of deriving a query pattern by deleting */? from the pattern string only worked for enumerated character classes. Only the query path opts in; document-content normalization is unchanged (insert/update/delete still strip wildcards, matching the TOKENIZE macro).
• fts_utils.cpp (tokenizeString): skip whitespace-only tokens that jieba CutForSearch emits between words. This surfaced once the query side actually used the configured tokenizer — e.g. the conjunctive query 深度学习 发展 produced a " " term that no document contains.
• fts_config.cpp: the custom ignore_pattern is now stored as-is for queries (ignorePatternQuery = ignorePattern) since wildcards are protected during normalization.
• New regression test fts/test/test_files/ignore_pattern.test reusing the existing fts-emails dataset (file names like allen-p/_sent_mail/102. carry digits/hyphens), covering exact queries, wildcard queries, and query behavior after index reload. No dataset changes needed.

Test plan

• Local: full FTS e2e suite passes (69 tests), including the new regression test
• CI