feat(cli): add --connect-timeout

[abonander]

And add a default 10 second connection timeout to all subcommands.

This addresses sporadic "connection refused" errors when using using sqlx-cli in automated jobs that start concurrently with a fresh database instance, such as using a Postgres image as a service in Gitlab CI.

Normally, I'd expect Gitlab to wait for the database container to be ready before proceeding with the build, and their documentation clearly states that this is already being done.

However, it appears that this is not the case for the Kubernetes executor which we're using in a self-managed Kubernetes cluster to speed up CI since builds with Gitlab's shared runners are glacially slow. Thanks to that misleading (or possibly outdated) bit of documentation, I wasted several hours today trying to figure this out.

Given that the above issue has been open for almost a year, I figured it was just easier to address it on our end.