Error hardening needed

[bastianjoel]

Currently errors like #210 or #197 result in an autoupdate service that does not serve any more updated data.

However - the service is still reporting to be healthy. This is problematic.
We need hardening against faults in the database layer. It is always possible that a query fails. This should not result in a broken service.

[bastianjoel]

When notify database connection is lost this issue happens again. This is now urgent.

[bastianjoel]

Moving to openslides-go as in none of our services it is handled that func (p *FlowPostgres) Update(ctx context.Context, updateFn func(map[dskey.Key][]byte, error)) returns.

That method should handle errors itself and never return unless the context is canceled.

[bastianjoel]

Fix will be improved as described in
#244 (comment)

[rrenkert]

@bastianjoel can this be closed?