Fix SQLInterpolator treating ? in comments and literals as placeholders + shorten next_changelog

[samikshya-db]

Summary

Fixes #1331 + some cosmetic change on next changelog

When supportManyParameters=1, SQLInterpolator.interpolateSQL split the SQL on every ? and counted every occurrence as a placeholder. Question marks inside line/block comments, single-quoted strings, double-quoted identifiers, and backtick-quoted identifiers were all counted, producing spurious "Parameter count does not match" errors for queries such as:

-- does this work?
select 'hello?', * from mytable where id = ?

This had 3 ? characters but only 1 real placeholder, so the driver rejected the perfectly valid binding of 1 parameter.

What changed

• Added SqlCommentParser.findPlaceholderPositions(sql) which walks the SQL with the existing comment/literal state machine and returns the source indices of ? characters that appear in State.NORMAL only.
• Reworked SQLInterpolator.interpolateSQL to use those positions: a single pass that slices the original SQL between placeholders and substitutes formatted parameter values. Comments, string literals, and quoted identifiers are preserved verbatim in the output.
• Removed the now-redundant countPlaceholders helper that scanned every ?.

DatabricksParameterMetaData.countParameters already used SqlCommentParser.forEachNonCommentChar and is unaffected.

Test plan

• Existing SQLInterpolatorTest cases (basic interpolation, mixed types, escaped quotes, mismatch errors, etc.) all pass.
• Added unit tests for SQLInterpolator covering ? inside line comments, block comments, single-quoted strings, double-quoted identifiers, backtick identifiers, and a combined case.
• Added unit tests for SqlCommentParser.findPlaceholderPositions covering null/empty input, basic positions, comments, literals, quoted identifiers, and escaped quotes.
• mvn spotless:apply clean.

[gopalldb]

Databricks SQL supports $$...$$ and $tag$...$tag$ dollar-quoted strings. The state machine doesn't track these. A ? inside $$...?...$$ would still be counted as a placeholder.

Looking at the code, the state machine only knows about ', ", and `. Dollar-quoting isn't supported.

Impact: If a user writes select $$hello?$$, ? from t with 1 param, the driver counts 2 placeholders → "Parameter count does not match".

Severity: Low — dollar-quoted strings aren't common in Databricks SQL. But the changelog and JavaDoc say "string literals" without qualification, which is misleading. Worth noting as a limitation, or adding
$$ support.

[samikshya-db]

Thanks for the suggestion, but I will just keep this out of scope for now.

[gopalldb]

Missing tests:

🟡 #13: No test for unclosed quote / comment (Corner Case #10).
🟡 #14: No test for surroundPlaceholdersWithQuotes with adjacent placeholders like select ?,? from t.
🟡 #15: No test for SQL containing only comments (-- foo\n with 0 params).
🟡 #16: No test for mixed escape inside identifier ( `col? ``) — the escape happens between ? characters.
🟡 #17: No test for interpolateSQL with extra parameters (params.size() > placeholders) — does it throw the right error?
🟡 #18: No test for very large SQL (perf regression check).
🟡 #19: ? immediately after -- start (--?). The -- enters IN_LINE_COMMENT; subsequent chars including ? are consumed in IN_LINE_COMMENT state. Should be skipped. Not tested directly.

[samikshya-db]

Thanks — added the ones I think are worth tracking:

• ✅ [PECO-948] Adds iterator for ArrowResultChunk #13 (unclosed quote/comment): added testFindPlaceholderPositionsUnclosedSingleQuote, …UnclosedDoubleQuote, …UnclosedBacktick, …UnclosedBlockComment — all assert 0 placeholders.
• ✅ [PECO-948] Read Data from Input Stream #14 (adjacent placeholders for surroundPlaceholdersWithQuotes): added a SELECT ?,?,? FROM t → SELECT '?','?','?' FROM t case to the parameterized test.
• ✅ [PECO-948] ArrowResultChunk Iterator implementation #15 (comments-only SQL): added testFindPlaceholderPositionsCommentsOnlySql covering line-only, block-only, and mixed comment-only inputs.
• ✅ [PECO-910] Implement parallel downloader for chunks as per configurable in-memory chunks #16 (mixed escape inside identifier): added testFindPlaceholderPositionsBacktickWithEscapeBetweenQuestionMarks with `a?b?`` — both?` chars stay inside the identifier.
• ✅ [PECO-996] Adding implementation for Prepared Statement #19 (--? immediately after line-comment start): added testFindPlaceholderPositionsLineCommentImmediatelyConsumesQuestionMark covering --?, --?\n, and --?\nselect ? from t.

Skipped:

• [PECO-948] Adds iterator logic to ArrowStreamResult and adds tests #17 (interpolateSQL with extra params) — already covered by SQLInterpolatorTest#testExtraParameters (line 64).
• [PECO-982] Changes to make the driver work with squirrel client for metadata #18 (very large SQL perf check) — skipping for now; perf assertions in JUnit tend to be flaky in CI, and a JMH micro-benchmark is the better venue if we ever want regression coverage.

[gopalldb]

Some minor comments, thanks for fixing this