fix(connection): remove heartbeat check in load balanced mode

[vkarpov15]

cc @alexbevi

Fix #15042
Revert #14812

Summary

We've gotten a few bug reports that indicate connectivity issues with Mongoose 8.7, including #14971 etc. Most notably, we've confirmed that #15042 is caused by the heartbeat check we added in #14812: we should not use that heartbeat check if the topology is LoadBalanced because the MongoDB Node driver doesn't send heartbeats in that case, so for LoadBalanced topology Mongoose currently loses connectivity after 20 seconds. Given this and all the issues we've gotten specifically referencing connectivity issues in Mongoose 8.7.0, I think it is prudent to remove this check for now.

Examples

[hasezoey]

Because #14812 main functionality is reverted, should we re-open #14727?

[vkarpov15]

Based on convo with @alexbevi I'm thinking it's better to make the following 2 changes:

1. Avoid the stale heartbeat check in LoadBalanced mode. Apparently the MongoDB Node driver doesn't receive heartbeats in LoadBalanced mode, and that's the mode the driver runs in when connected to MongoDB behind haproxy or when connected to Atlas serverless instances.
2. Don't treat successful connection as a heartbeat. Part of the problem here is that Mongoose thinks it received a heartbeat because we set _lastHeartbeatAt when Mongoose successfully connects, but then no other heartbeats are ever received. With this change, if Mongoose never received an actual successful heartbeat event, it won't use the stale heartbeat check. This change is a precaution for other edge cases where Mongoose might not receive any heartbeats for some reason.