Links with nested HTML or \n confuse the markdown converter

A couple cases that generate the wrong markup.

### 1. nested html elements in link
```html
foo bar foo bar
```
gets converted to

```markdown
[foo](http://example.com/) bar foo [bar](http://example.com/)
```

when I would expect

```markdown
[foo](http://example.com/) [bar](http://example.com/) foo bar
```

What happens: the walker enters the `a` and then enters the `span` again, finding the same text twice. But the proper label text was already consumed, so it consumes until it finds another match. In some cases (like above) it even linkifies the wrong text.

### 2. line breaks in link
```html
foo
bar
```
(notice the line break) gets converted to

```markdown
foo bar
```

when I would expect

```markdown
[foo bar](https://example.com/)
```

What happens: the node textContent is `'foo\nbar'` which doesn't match any text in the plaintext.

### 2b. `
` in link
Another similar (but possibly harder to fix) case is

```html
foo
bar
```

which like 2 doesn't create a link:

```markdown
foo
bar
```

---

In general, since the code can't rely on the browser to properly deal with HTML content, some of these corner cases will probably keep popping up. But this happened in the real world (try copying and pasting the first news entry in the deprecated section [here](https://docs.python.org/dev/whatsnew/3.10.html#deprecated)) and it seemed to be significant enough to report.

I will follow up with a PR that suggests potential fixes for 1 and 2 (but not 2b), but this is not my field so it might be far from good.

The issue names no files or tests. Start by reproducing the three HTML-link cases from the report in the markdown converter, then inspect the anchor-text matching and node textContent handling described there. Done means nested elements and line breaks in links produce the expected Markdown links, with regression coverage for the reported examples.

由索引模型根据 Issue 内容生成。