perf(p2p): Remove unnecessary atomic read

[ValarDragon]

This PR is a driveby change. We were doing an atomic read here, but this is unnecessary. Nothing in the codebase modified this variable after instantiation.

Doesn't really feel changelog worthy, but I can add one if we want. Felt like a minor code nit

[melekes]

Thanks @ValarDragon ❤️

[sergio-mena]

Hi, getting here a bit late...

I've checked the call hierarchy of sendSomePacketMsgs and I see that, ultimately, it's being called from two different goroutines (potentially for the same MConnection):

• It's called from MConnection.FlushStop <-- pex.Reactor.Receive (actually, an ad-hoc goroutine there)
• It's called from MConnection.sendRoutine, which is run as a goroutine

In principle, we could suspect there might be a race between FlushStop and sendRoutine. However, FlushStop waits on channel MConnection.doneSendRoutine before calling sendSomePacketMsgs, so we should be good.

Posting this here to further document this change.

[cason]

Just for context, calling FlushStop more than once is not a problem. The method startswith a call to stopServices. This method is protected by a lock and once the first execution of this method completes, all the following will return true. When this method returns true, the FlushStop method is a no-op.