fix: add divider row to markdown tables with auto-generated integer headers

[vibeyclaw]

Problem

Fixes #90

When an HTML table has no semantic <thead>, pd.read_html() assigns integer column names (0, 1, 2, ...). The previous code stripped both the integer header row and the separator row, producing output like:

| Revenue | 2024 |
| 5.2B    | 2023 |
| 4.8B    | 2022 |

This is invalid Markdown — no divider row means the table is not rendered as a proper table in GitHub, Obsidian, or other Markdown renderers.

Fix

After stripping the auto-generated integer headers, promote the first data row to serve as the visual header and insert a generated --- divider beneath it:

| Revenue | 2024 |
|---------|------|
| 5.2B    | 2023 |
| 4.8B    | 2022 |

The elif branch (for tables with meaningful column names) is unchanged — those already produce a valid divider.

Root Cause

Many SEC filing tables use <td> for both header and data rows without a <thead> element, so pandas cannot distinguish the visual header from data. The fix preserves the original intent of stripping integer placeholders while ensuring the output is always valid Markdown.