network: fix mesher stopping logic

[algorandskiy]

Summary

Originally Mesher relied on a supplied context to handle termination so network's context supposed to be canceled first. wsNetwork's meshThreadInner executes tryConnect asynchronously so that calling it directly from Mesher after possible network termination ends up with tryConnect goroutine still running.

Sample failure: https://github.com/algorand/go-algorand/actions/runs/16656257238/job/47141871869?pr=6397

[cce]

I am curious, why make a child context and cancel it in baseMesher.stop() if the parent context will be cancelled by the network upstream? or is the issue that for hybridNetwork there is no WithContext() being passed or something

[algorandskiy]

the issue with hybrid is the following (in the original code):

1. Stoping wsNet and then the mesher
2. In between wsNet.Stop() and mesher.stop() mesher's goroutine calls directly meshThreadInner that spawns a goroutine for tryConnect
3. Since wsNet.Stop() has already executed, nothing checks for tryConnect wait group and the goroutine still alive when the main thread exits.