MOD-14850 Fix crash in race between MR_ClusterAsyncDisconnect and MR…

[AvivDavid23]

…_ClusterOnDisconnectCallback
Fix a race in MR_ClusterAsyncDisconnect: a deferred disconnect task scheduled after a TLS handshake failure runs after MR_ClusterOnDisconnectCallback has already nulled n->c, causing redisAsyncFree(NULL).

---

Note

Medium Risk
Touches async connection lifecycle and TLS error handling, where subtle ordering issues can cause regressions; changes are small and covered by new race/TLS tests.

Overview
Prevents a crash when a deferred MR_ClusterAsyncDisconnect runs after hiredis has already processed a disconnect and cleared n->c (common on TLS handshake failures) by guarding redisAsyncFree and nulling the pointer after freeing.

Cleans up the TLS connect failure path by removing a temporary hiredis callback-reset workaround, and extends the test suite with a targeted reproduction (test_disconnect_race.py), TLS-capable shard mocks in test_network.py, and a single-node TLS test run in the test module Makefile.

^{Reviewed by Cursor Bugbot for commit 5e13e85. Bugbot is set up for automated code reviews on this repo. Configure here.}

[AvivDavid23]

This pr also delete some temporary fix code which is not needed anymore(#51)