fix(postgres): avoid unnecessary flush in PgCopyIn::read_from

[jwangnz]

This pull request addresses a potential hang in PgCopyIn::read_from caused by conn.stream.flush().await?;. The flush call appears to be unnecessary at this point, as the write buffer has already been flushed in PgCopyIn::begin.

[abonander]

I agree that the flush is arguably unnecessary but I'm more interested in why it would hang. It should return immediately if there's nothing to flush.

Do you have a reliable reproduction for this?

[abonander]

Are you using rustls or native-tls?

[jwangnz]

@abonander we are using rustls, and unfortunately I don't have way to reproduce it yet. We have three environments using the same binary with the same version of AWS RDS for Postgres, but only one has this issue.