[MM-63434] Use forked PDF library with parsing depth limit (cherry-pick #35947) (#36177)

* [MM-63434] Use forked PDF library with parsing depth limit (#35947)

* [MM-63434] Use forked PDF library with parsing depth limit

Replace github.com/ledongthuc/pdf with a fork that limits object
nesting depth during parsing. Add test coverage.

* Reverting incorrect merge that lost the change to msgpack

The error was in merge 64bdff88d8

* Remove stale ledongthuc/pdf checksums after fork replace (go mod tidy)

* Fix TestPdfFile expected text for forked PDF extractor (release-10.11)

The jgheithcock/pdf fork returns a leading newline for sample-doc.pdf; align
with master and 11.x cherry-picks.

server/platform/services/docextractor/pdf_test.go

@@ -20,7 +20,7 @@ func TestPdfEmptyFile(t *testing.T) {
 
 func TestPdfFile(t *testing.T) {
 	extractor := pdfExtractor{}
-	contentText := "This is a simple document that contains some text."
+	contentText := "\nThis is a simple document that contains some text."
 	content, err := testutils.ReadTestFile("sample-doc.pdf")
 	require.NoError(t, err)
 	extractedText, err := extractor.Extract("sample-doc.pdf", bytes.NewReader(content), 0)
@@ -28,6 +28,21 @@ func TestPdfFile(t *testing.T) {
 	require.Equal(t, contentText, extractedText)
 }
 
+func TestPdfDeeplyNestedObjects(t *testing.T) {
+	// Test for MM-63434
+	var buf bytes.Buffer
+	buf.WriteString("%PDF-1.0\n")
+	for range 10_000 {
+		buf.WriteString("0\n0\nobj\n")
+	}
+	buf.WriteString("startxref\n0\n%%EOF\n")
+
+	extractor := pdfExtractor{}
+	text, err := extractor.Extract("excessive-nests.pdf", bytes.NewReader(buf.Bytes()), 0)
+	require.Error(t, err)
+	require.Empty(t, text)
+}
+
 func TestWrongPdfFile(t *testing.T) {
 	extractor := pdfExtractor{}
 	content, err := testutils.ReadTestFile("sample-doc.docx")