Add automated checks for text cleanup and tokenization behavior

[CocoHusky]

Context

Rapid Reader depends on deterministic text cleanup, simplification, tokenization, punctuation pauses, and long-word rendering behavior. The README has a strong manual verification checklist, but these text-processing rules are good candidates for lightweight automated tests.

Proposed work

• Add focused tests for cleanup rules such as code block replacement, inline code replacement, URL replacement, noisy symbol removal, link removal, hyphen splitting, and numeric citation stripping.
• Add tests for token counts and edge cases like empty documents, long words, punctuation-heavy text, and Markdown tables.
• Document the test command in the README if it differs from the existing build/check flow.

Acceptance criteria

• Core cleanup/tokenization behavior can be verified without opening Obsidian.
• Manual verification remains focused on UI behavior.
• The tests do not require network access or a real vault.