Implement server side connection limit

[jbearer]

This change implements a configurable limit on the number of simultaneous HTTP connections allowed by a query server. This should allow us to efficiently filter out requests during periods of high load so that the burden is pushed back onto clients, and the server continues operating within normal limits. This in turn will allow us to turn off or scale back the AWS rate limiter.

This PR:

• Uses the RateLimitListener from the new version of tide disco to limit server load to a configurable number of requests

This PR does not:

• Rate limit based on client IPs. Rate limiting is first come first served, so we will still need an intelligent, external layer to prevent malicious behavior and DDoS attacks. But the parameters of this external rate limiter can be tuned to handle deliberate attacks only, while heavy load due to normal circumstances should be adequately handled by the server
• Limit connections open by WebSockets. The limit only applies to HTTP connections

Key places to review:

• sequencer/src/api/options.rs

Things tested

Set up local demo to use a fairly aggressive limit of 25 simultaneous requests. Ran just demo-native and the logs of the load generators, nasty-client, and sequencers were all happy.