fix(drivers): show pagination for SELECTs with leading SQL comments

[ymadd]

Summary

• is_select_query now strips leading -- / /* */ comments before checking for SELECT, matching returns_result_set and is_explainable_query. Without this, a SELECT preceded by a header-comment block was routed through the result-set fetch path but bypassed the pagination branch in exec_on_*_conn, returning pagination: None and hiding the UI controls entirely — so a 500-row result looked like a fixed slice with no way to advance.
• Fixed a latent bug in strip_limit_offset that the classifier change exposed: it tokenised then rejoined with " ", collapsing newlines. For a query like -- header\nSELECT ... LIMIT 50 the rejoined output became -- header SELECT ... LIMIT 50 OFFSET 0, which MySQL/Postgres/SQLite all parse as a single -- comment line — silently no-op. Switched the tokenizer to track byte offsets and made strip_limit_offset slice the original input, preserving comments and whitespace verbatim.
• Factored the three identical quoted-token scanner branches (', ", `) into a read_quoted helper while refactoring the tokenizer.

Repro (before this PR)

-- ============================================
-- Any header comment block
-- ============================================
SELECT * FROM big_table ORDER BY id;  -- 500+ rows

Result: rows show, but the pagination bar is gone — only the first page is reachable.

After: pagination renders and pages can be navigated as expected.

Test plan

• cargo test --lib drivers::common — 42 passed, including 4 new regression tests:
  • test_is_select_query_with_leading_comments
  • test_build_paginated_query_with_leading_comments
  • test_build_paginated_query_multiline_comments_with_user_limit
  • test_strip_limit_offset_preserves_leading_comments
• Manual: paste a comment-headered SELECT returning >page_size rows against MySQL, Postgres, SQLite; verify pagination bar renders and Next/Prev work.

🤖 Generated with Claude Code