HTML tables lose the header delimiter row after `<caption>` or `<colgroup>`, and `<tfoot>` adds a fake header

When a table's rows sit directly under <table> (not in <thead> or <tbody>), a <caption> or <colgroup> before them drops the | --- | line under the header. Without it the output isn't a Markdown table anymore, just lines of pipes. JDK 8 and 11 Javadoc summary tables are built this way.

A <tfoot> row has the opposite problem. It gets its own empty header and | --- | line in the middle of the table.

Repro on main (945314a, markdownify 1.2.3):

from markitdown.converters import HtmlConverter

c = HtmlConverter()
print(c.convert_string(
    "<table><caption>Stock</caption><tr><th>Name</th><th>Qty</th></tr>"
    "<tr><td>Apple</td><td>3</td></tr></table>").markdown)
print(c.convert_string(
    "<table><thead><tr><th>Item</th><th>Cost</th></tr></thead>"
    "<tbody><tr><td>A</td><td>1</td></tr></tbody>"
    "<tfoot><tr><td>Total</td><td>1</td></tr></tfoot></table>").markdown)

What I get:

Stock

| Name | Qty |
| Apple | 3 |

| Item | Cost |
| --- | --- |
| A | 1 |
|  |  |
| --- | --- |
| Total | 1 |

I expected | --- | --- | under the first header, the same as without the caption. The footer should just be the last row, | Total | 1 |.

Start at the HtmlConverter entry point and reproduce both cases with convert_string using the HTML snippets in the issue. Trace the table conversion handling for caption, colgroup, and tfoot. Done means captioned or colgroup tables retain the Markdown delimiter row, while tfoot contributes only its data row without creating another header.