RFR 8232733: Remove need to grab Threads_lock while processing handshakes

[Robbin Ehn]

Hi Patricio,

> One reason was so that the JavaThread would not be freed while 
> reading it (although looking at that first commit it seems that between releasing the Threads_lock after setting the 
> handshake and acquiring it again before processing it, the JavaThread could have already processed the handshake and 
> exited). 

My memory say this would have been okay, that thread would not been seen any more on threads-linked-list.
Thus next iteration VM thread wouldn't see it, but still wait for done semaphore.

Anyhow, I think this is great removing Threads_locks.

Looks good, thanks, Robbin.

> 
> Thanks,
> 
> Patricio
>> Thanks,
>> David
>>
>>
>>> Thanks,
>>>
>>> Patricio
>