reverseproxy: SRV dynamic upstream failover

[mholt]

This should fix #5816 - if the SRV lookup fails and a GracePeriod is defined, it will continue to use the previously-cached records until the GracePeriod is up, after which it will try again.

@cds2-stripe @jjiang-stripe Feel free to give this a shot -- let me know if I need to rebase this onto another branch.

[kkroo]

Thanks for this fix @mholt I’m wondering why it’s only for SRV and not A lookups for dynamic upstreams?

[mholt]

It was only requested for SRV upstreams for the moment. If it for sure works we can easily add it for A upstreams too :+1:

[francislavoie]

This is missing Caddyfile support for the grace_period option