fix: preserve Markdown hard line breaks

[Maciej Zmitrukiewicz (MaciejZet)]

Summary

• MarkItDown._convert was stripping every trailing space, so a CommonMark hard break (two or more spaces before a newline) became a soft break.
• Lines that end with at least two spaces are now kept. A single trailing space, a trailing tab, and a whitespace-only line are still removed. Runs of blank lines are still collapsed.
• HTML <br> and DOCX <w:br/> are emitted by markdownify as those two spaces, and current upstream expectations now preserve the hard break.

The branch has been rebased onto the current upstream main to resolve the merge conflict. The fix is limited to the normalization helper and its regression coverage; unrelated upstream test refactors were preserved.

Test plan

• Full packages/markitdown/tests on Python 3.12 with the all extra: 1124 passed, 12 skipped
• Dedicated hard-break regression: 29 passed
• Updated HTML and DOCX expectations for generated hard breaks

Refs #2559

[Maciej Zmitrukiewicz (MaciejZet)]

@microsoft-github-policy-service agree