Fix writer nullpointer panic on network reconnect

[SijmenHuizenga]

In #655 the change was made in client.go to set writer to nil, before completing the 'closing' of the client. This makes that if someone calls WritePacketRTPWithNTP after writer = nil but before c.done is closed, there is a panic. As described in #678.

I don't fully understand why c.writer = nil is needed in the first place. It seems that before #655 c.writer is never set to nil. And after testing the situation of the panic, this resolved it. If someone has a better solution I'm very eager to learn!

[aler9]

writer was set to nil to prevent chWriterError from returning a closed channel and therefore messing up the client loop. Anyway, this was causing issues both when the client was closing and when the client was paused, so the fix is right. Thanks.

[SijmenHuizenga]

AMAZING! Thank you so much for the proper fix!