[Connector] Do not fail job when PulsarAdmin API fails when getting the data.

Currently, Pulsar connector periodically queries the admin api to detect partition changes. However, this can cause the pipeline to fail if the admin api fail. 
@weixiangchen Reported a admin 502 when running pipeline for longer peirods (30 minutes).

We should discuss whether we want to 
1) fail the job
2) restart the job 
3) report the exception and go on

when similiar 502 exception happens. 

