fix(stack): surface fetch_old_pr_heads failures instead of swallowing

The exception handler around `fetch_old_pr_heads` previously did a
silent `pass`, leaving every revision-history entry stamped
"unknown" with no clue why. When a real failure happens (fork remote
without `refs/pull/N/head`, network error, missing permissions, …)
the user has no signal that anything went wrong — they only see the
symptom in the PR comment days later.

Log the underlying error in orange so the cause is visible at push
time. Behaviour stays non-fatal: classification still falls back to
"unknown" and the push proceeds.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Change-Id: Id33459ea069be228f73fb02a7f813f16db145b71

Author: jd

mergify_cli/stack/push.py

@@ -589,8 +589,13 @@ async def stack_push(
             with console.status("Fetching old PR heads for comparison..."):
                 try:
                     await fetch_old_pr_heads(remote, updated_pr_numbers)
-                except utils.CommandError:
-                    pass  # Non-fatal: change type will be "unknown"
+                except utils.CommandError as exc:
+                    # Non-fatal: change type will be "unknown" — but surface
+                    # the underlying error so the user can fix it.
+                    console.log(
+                        f"[orange]Could not fetch old PR heads; revision-history "
+                        f"change types will fall back to 'unknown': {exc}[/]",
+                    )
 
             # Detect change types before force-push overwrites refs
             change_types: dict[str, str] = {}