IPv6/Multi: Let IP-task use vSocketCloseNextTime instead of FreeRTOS_closesocket

[htibosch]

Description

This PR does the same in the IPv6/multi branch as PR #238 did for the IPv4 branch.

Background: the function FreeRTOS_closesocket() can fail if the queue to the IP-task is full. A normal task can wait for space, but the IP-task would create a dead-lock by waiting for itself.

This PR has a safe alternative for FreeRTOS_closesocket() that can only be used by the IP-task.

Test Steps

I tested this PR by corrupting the TCP flags of an incoming packet during the SYN phase around here. That leads to an immediate closure of the socket like this:

    /* In case pxSocket is not yet owned by the application, a closure
     * of the socket will be scheduled for the next cycle. */
    vTCPStateChange( pxSocket, eCLOSE_WAIT );

Related Issue

It is recommended to make the message queue big enough so that it never gets full. Just like a stack should never overflow.
The queue length is defined as ipconfigEVENT_QUEUE_LENGTH. I often define it as:

#define ipconfigEVENT_QUEUE_LENGTH   ( ipconfigNUM_NETWORK_BUFFER_DESCRIPTORS + 5 )

See also uxGetMinimumIPQueueSpace(), which is defined when
ipconfigCHECK_IP_QUEUE_SPACE=1. It can be used to monitor the queue size.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.