Timestream's `ReloadEndpoint` is unclear

### Describe the issue

I've made good attempts to understand how Timestream's ReloadEndpoint works, which doesn't seem to documented well. It still doesn't make sense to me, unfortunately.

The docs.rs of [`ReloadEndpoint`](https://docs.rs/aws-sdk-timestreamwrite/latest/aws_sdk_timestreamwrite/endpoint_discovery/struct.ReloadEndpoint.html) only says "Endpoint reloader."

Here's roughly our code:
```rust
let config = aws_config::defaults(aws_config::BehaviorVersion::latest())
.region(REGION)
.load()
.await;
let (client, reload) = aws_sdk_timestreamwrite::Client::new(&config)
.with_endpoint_discovery_enabled()
.await
.unwrap();

// This task will terminate when the corresponding Client is dropped.
tokio::task::spawn(reload.reload_task());

... work with the client here.
```

My questions are:
- What function does the `ReloadEndpoint` have?
- Why do we need to call `.with_endpoint_discovery_enabled()`? It will not work without this.
- It seems to work without me spawning the reload task - What does that task really do then?

### Links

https://github.com/awslabs/aws-sdk-rust/blob/29883b151225f7b27a22e4d343585d2dc8207521/sdk/timestreamwrite/src/client.rs#L18-L23

Start with sdk/timestreamwrite/src/client.rs at the linked endpoint-discovery code, then compare it with the docs.rs ReloadEndpoint documentation. Trace how with_endpoint_discovery_enabled and reload_task are used, and document the reloader's purpose, why opt-in is required, and what spawning the task does. Done means the three questions are answered clearly for users.

Written by the indexing model from the issue text.