Fix commonmark regression test to check the FileId it actually protects

roborev review (job 1730) caught that commonmark_reader_source_context_uses_forward_slashes
asserted FileId(0), but ASTContext::with_filename populates FileId(0)
internally before commonmark::read's own add_file call ever runs — the
entry that call actually writes is FileId(1). The test passed regardless
of whether that call normalized its filename, giving zero protection for
the fix in f5a93fa. Verified by reverting just that one line and
confirming the test still passed; asserting FileId(1) instead now fails
correctly when reverted.

Author: cderv

crates/pampa/tests/integration/test_diagnostic_path_normalization.rs

@@ -102,9 +102,14 @@ fn commonmark_reader_source_context_uses_forward_slashes() {
     let (_, context) =
         pampa::readers::commonmark::read("hello\n", "tests\\snapshots\\commonmark\\001.qmd");
 
+    // FileId(0) is the entry `ASTContext::with_filename` creates internally
+    // (already normalized, unrelated to this reader's own fix). The fix
+    // this test protects is commonmark::read's own second `add_file` call,
+    // which lands at FileId(1) — asserting FileId(0) here would pass
+    // regardless of whether that second call normalizes its filename.
     let file = context
         .source_context
-        .get_file(FileId(0))
-        .expect("file 0 should exist");
+        .get_file(FileId(1))
+        .expect("file 1 should exist");
     assert_eq!(file.path, "tests/snapshots/commonmark/001.qmd");
 }