optimize the calculation method of ticdc's gc saftpoint

[nongfushanquan]

Is your feature request related to a problem?

As for one upstream in lots of changefeeds scenarios, if one changefeed fails and other changefeed work normal, ticdc will ignore the checkpoint ts of the failed changefeed and use the the checkpoint ts of the normal one as gc safe point. So the tidb would gc some data before the customer handle the error.

Describe the feature you'd like

if one changefeed fails, the gc safe-point can hold for 24h . The calculation method is as followed:

If changefeed1: failed （check point ts = ts1)，changefeed2 normal checkpoint ts = ts2.
Then Gc safepoint = min( ts2, max( ts1, currentPDts - 24h))

Describe alternatives you've considered

No response

Teachability, Documentation, Adoption, Migration Strategy

No response

[nongfushanquan]

/area ticdc

[nongfushanquan]

/assign @charleszheng44

[nongfushanquan]

/label affects-7.1

[nongfushanquan]

/label affects-6.5