Make --progress and --checkpoint strictly by-statement

[rolandwalker]

Description

Previously --progress and --checkpoint were influenced by linebreaks to some extent: multiline queries were correctly joined and counted/dispatched/checkpointed as one query, but multiple queries on a single line were dispatched together.

That means that the progress estimation could be thrown off somewhat, depending on the file contents, and more importantly means that a statement which was part of line with more than one statement might fail to be written to the line-influenced checkpoint file if that particular query succeeded, but a subsequent query on the same line failed.

This subtlety is important if we are to use the checkpoint file to resume scripts, though in general it would be best when running scripts to avoid all of these corner cases by having one statement per line.

We pull in sqlparse in addition to sqlglot, because sqlparse has the feature of preserving the input literally when splitting multi-statement lines.

This also fixes a bug: the generator here named batch_gen was recreated in the --progress loop, which didn't matter before this change since iterating over a filehandle covered up the issue.

Tests are added for statements_from_filehandle(), which had no coverage before.

Incidentally

• fix missing changelog entry
• fix whitespace in a comment
• remove a backslash by double-quoting a string which contains a single quote

Checklist

• I added this contribution to the changelog.md file.
• I added my name to the AUTHORS file (or it's already there).
• To lint and format the code, I ran

  uv run ruff check && uv run ruff format && uv run mypy --install-types .

[github-actions]

No correctness or security issues stood out in the PR-scoped changes.

Residual risk / test gap:

1. I could not run the targeted tests locally because uv is not available in this environment (uv: command not found), so this review is static-only.
2. Coverage is improved for statement splitting and --progress, but there is still no end-to-end main.py test for --batch (without --progress) with multiple statements on one line; adding one would better guard regression in the non-progress path.