test: expand GoJieba fulltext BVT coverage

[Ariznawlll]

What changed

• add focused GoJieba tokenizer BVT coverage for exact tokens, UTF-8 positions, DocLen, mixed text, repeated terms, token-length limits, and multi-column boundaries
• cover the remaining SQL-level tokenizer gaps: dictionary-only/HMM-disabled segmentation plus whitespace-only, punctuation-only, empty, and NULL input
• cover NULL/empty-value indexing, duplicate-term and cross-document negatives, empty boolean-pattern rejection, non-indexed-column updates, and TRUNCATE/reinsert visibility
• run the deterministic DISTINCT, GROUP BY, HAVING, CTE, derived-table, and ordinary-predicate query shapes through a GoJieba index
• retain unsupported query shapes as standard issue-tagged expected failures

Why

These small deterministic contracts were primarily exercised by nightly regression, so regressions could be detected later than necessary. The new BVT cases move fast functional coverage into the MatrixOne repository without duplicating existing JOIN, UNION, ranking, JSON, primary-key update, large-data, or transaction scenarios.

Impact

Test-only change. No production code or transaction cases are modified.

Validation

• mo-tester: 53 total, 51 passed, 2 expected issue skips, 0 failed, 0 abnormal
• go test ./pkg/sql/plan -run FullText -count=1
• git diff --check

Closes #25896

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[gouhongshen]

Codex automated review

One non-blocking BVT determinism issue found. Tokenizer unit tests pass; broader tests were blocked by missing local C headers.

P2 - Order exact tokenizer results explicitly (test/distributed/cases/fulltext/gojieba_precise.sql:10)

The six select f.pos, f.word queries assert a specific row sequence, including __DocLen last, but have no ORDER BY. SQL does not guarantee table-function result order, so plan or distributed-execution changes can make this BVT flaky without changing tokenization. Add ORDER BY f.pos to each exact-token query.