ErrGCTTLExceeded is not a fast-fail error

[liuzix]

What did you do?

Start a CDC cluster with server config gc-ttl = 600 (10 minutes). Accumulate some data upstream to produce the scenario where checkpoint-ts is too behind.

What did you expect to see?

We expect the changefeed to fail.

What did you see instead?

The changefeed retries instead of failing.

Versions of the cluster

Upstream TiDB cluster version (execute SELECT tidb_version(); in a MySQL client):

v5.1.2

[asddongmen]

wip