slack-19.0: support consul topo stale reads

[timvaillancourt]

Description

This PR allows Vitess components to use stale reads from consul. This is enabled with the flag --topo_allow_stale_reads, which is default false and should be considered experimental

*api.QueryOptions:

// AllowStale allows any Consul server (non-leader) to service
// a read. This allows for lower latency and higher throughput
AllowStale bool

This flag allow us to distribute topo .Get(...) and .List(...) load to consul replicas. Also it limits the impact of elections

It may be safe in use cases like txthrottler if stale reads have no impact to any other vttablet operations. It may be safe to use in vtorc also

Related Issue(s)

Checklist

• "Backport to:" labels have been added if this change should be back-ported
• Tests were added or are not required
• Did the new or modified tests pass consistently locally and on the CI
• Documentation was added or is not required

Deployment Notes

[ejortegau]

LGTM but some CI failed. I triggered a re-run to see if that helps

[timvaillancourt]

LGTM but some CI failed. I triggered a re-run to see if that helps

@ejortegau the query serving downgrade tests on v19 are known-broken and currently not required CI

For context, we've confirmed we shouldn't be impacted by the handful of failures in those CIs but we still need to modify the pipeline to understand this

[github-actions]

This PR is being marked as stale because it has been open for 30 days with no activity. To rectify, you may do any of the following:

• Push additional commits to the associated branch.
• Remove the stale label.
• Add a comment indicating why it is not stale.

If no action is taken within 7 days, this PR will be closed.