LinkReferenceDefinition source position is ignored in NormalizeRenderer

the test is follow:

type MarkdigTest(output: ITestOutputHelper) =
    let text =
        Path.Combine(@"C:\Program Files\Typora\resources\app\Docs","Markdown Reference.md")
        |> File.ReadAllText

    let document =
        Markdown.Parse(text, MarkdownPipelineBuilder().UseAdvancedExtensions().Build())

    [<Fact>]
    member this.``NormalizeRenderer Test``() =
        use writer = new StringWriter()
        let normalizer = new NormalizeRenderer(writer)
        normalizer.Render(document) |> ignore
        let raw = writer.ToString()
        File.WriteAllText(@"d:\Markdown Reference.md",raw)

the test file is, need remove the .txt extent name:
Markdown Reference.md.txt

please note that Footnotes, Reference Links, URLs, this is normalized file:
Markdown Reference.md.txt

Start with the F# reproduction in the issue and inspect NormalizeRenderer, comparing its output with the linked Markdown Reference.md examples. Determine how LinkReferenceDefinition source positions are handled during normalization, then add a focused regression test showing the corrected output.