Fix simultaneous reconnection handling and exponential reconnection interval

[jaunis]

Hello,

I ran into an issue in which a 'close' and an 'error' event are fired simultaneously. Thus, 2 re-connections are launched at the same time, and the second removes the listeners positioned on the socket by the first one.
This causes errors raised by the newly created socket not to be caught.

In this pull request, I fix this issue by rejecting a reconnection request if another one is pending.

I also fix the exponential reconnection interval : it is now limited to five times the initial interval.

Regards

[andrew-aladev]

Please add a comment about Sigmoid function you used.

[andrew-aladev]

Please use retry.pending or something similar instead of instance variable. It looks like @nucleardreamer is designed retry object for such usage.

[jaunis]

@andrew-aladev It's done

[nucleardreamer]

Thanks for the commit @jaunis, and sorry about the radio silence. I will be able to review this later today