fix: preserve HTML tables in Outlook .msg conversion

[octo-patch]

Fixes #1567

Problem

The OutlookMsgConverter reads only the plain text body, discarding the HTML body. When a .msg file contains HTML tables, the plain text fallback strips all HTML formatting.

Solution

Prefer the HTML body (PR_BODY_HTML) when it exists:

1. Try Unicode HTML stream __substg1.0_1013001F first
2. Fall back to binary HTML stream __substg1.0_10130102
3. Convert HTML to markdown via BeautifulSoup + _CustomMarkdownify (same as HtmlConverter)
4. Fall back to plain text body if no HTML body is present

No new dependencies introduced.

Testing

Tested with a .msg file containing HTML tables. Before: unformatted plain text. After: proper markdown tables.

[VANDRANKI]

The priority order (HTML body first, plain text fallback) is correct - MSG files that have an HTML body should produce better markdown than stripping it to plain text.

A few points:

script/style removal - good. Without this, CSS and JS in the HTML body would end up in the markdown output.

assert olefile is not None in _get_binary_stream_html - the existing _get_stream_data method uses assert olefile is not None too, so this is consistent with the existing pattern. But both methods are only called after the class-level import check, so the assertion should never fire in practice - it is really just a type-narrowing hint. A comment explaining this would help future readers.

BeautifulSoup dependency - is this already available in the optional deps for OutlookMsgConverter? If not, it needs to be added to the extras in pyproject.toml alongside olefile.

Tests - is there a test .msg file with an HTML body that exercises the new path? If not, a test covering table preservation would strengthen this.

The logic itself looks correct.