Keep weak pointer to node in socket objects

[pwojcikdev]

There were occasional lifetime issues when running tests, where socket handlers would complete after parent node is destroyed. While using weak ptr to fix such issues is arguably a symptom of bad design, it's a necessary intermediate step to continue refactoring network code to eventually have async handlers properly controlled.

[clemahieu]

Looks like a simple mechanical change and I don't see any behaviour changes besides the weak_ptr.