fix(richtext-lexical): combine 2 normalizeMarkdown implementations and fix code block regex (#10470)

This should fix it https://github.com/payloadcms/payload/issues/10387

I don't know why we had 2 different copies of normalizeMarkdown.

Also, the most up-to-date one still had a bug where lines were
considered as if they were inside codeblocks when they weren't.

How I tested that it works:

1. I copied the `normalizeMarkdown` implementation from this PR into the
website repo, and made sure it is called before the conversion to
editorState.
2. In the admin panel, sync docs.
3. In the admin panel, refresh mdx to lexical (new button, below sync
docs).
4. Look for the examples from bug #10387 and verify that they have been
resolved.

An extra pair of eyes would be nice to make sure I'm not getting
confused with the imports.

test/lexical-mdx/int.spec.ts

@@ -174,8 +174,6 @@ describe('Lexical MDX', () => {
           ? (sanitizedInputAfterConvertFromEditorJSON ?? sanitizedInput).replace(/\s/g, '')
           : (sanitizedInputAfterConvertFromEditorJSON ?? sanitizedInput)
 
-        console.log('resultNoSpace', resultNoSpace)
-        console.log('inputNoSpace', inputNoSpace)
         expect(resultNoSpace).toBe(inputNoSpace)
       })
     }