Make prettify/unprettify logic more robust

[rolandwalker]

Description

• explicitly handle more exceptional cases such as empty input
• clarify in error message that only a single statement is expected
• don't return an empty string on failure; instead return the original text

Works together with #1645.

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]

1. Potential regression: failure path mutates input instead of returning original text
   In both methods, parse failure now does text.rstrip(';') rather than returning the original text. This strips all trailing semicolons and can produce '' for inputs like ';'/';;;', which conflicts with the PR goal (“return the original text on failure”).

   • mycli/main.py:857
   • mycli/main.py:873
     Action: on failure, return text unchanged; manage semicolon appending only in the success branch.

2. State-coupling bug risk: semicolon append depends on global error state
   Successful parse appends ; only when not self.toolbar_error_message. If an earlier failure left this flag set, a later successful prettify/unprettify can skip semicolon append unexpectedly.

   • mycli/main.py:859
   • mycli/main.py:875
     Action: use a local parse_failed flag (or clear toolbar_error_message at method start / in success branch) so success formatting is independent of stale UI state.

Missing tests / edge cases

• Add tests for failure-path exact preservation of input (';', ';;;', 'SELECT 1;;', unparsable SQL).
• Add test for “failure then success” on same MyCli instance to ensure success still appends semicolon and error state is reset appropriately.

No direct security concerns found in this PR diff.
I did not run tests here because required runtime deps (e.g., sqlglot) are not available in this environment.

[rolandwalker]

Merging this without review, since it is pretty much required for #1645, which I merged without noticing the state of this one.