Report reply's --all-identical Batch Count When a GraphQL Call Exits the Process #2421

What happens

reply --all-identical answers each identical thread in turn. When one copy's reply or resolve fails at the transport level, or returns a GraphQL error, gh_graphql raises SystemExit. reply_to_thread doesn't catch it. The process exits with gh graphql failed and prints no batch count. Copies earlier in the batch may already carry the reply and be resolved, and nothing tells the operator how many did.

The batch line N of M identical threads are confirmed to carry this reply only prints on the 62 and 63 exits.

Shape that reproduces it

• Three identical unresolved threads on one file.
• The first copy's reply and resolve succeed.
• The second copy's addPullRequestReviewThreadReply call fails, for example with a network error.
• The output ends at gh graphql failed, and no count is printed.

Where

scripts/pr_review.py reply_to_thread's batch loop, and gh_graphql's SystemExit on failure. The local strict review of #2420's follow-up fix found this. It predates the count change.