Always use the passed logger (from Messenger) in msi_kill.

[iphydf]

Narrowing down #1149.

---

This change is 

[nurupo]

Aren't they the same logger anyway or do we actually have multiple logger instances per a toxcore+toxav instance?

[nurupo]

Oh, is this because session->messenger might no longer exist at this code path?

[nurupo]

Looking at that function, it assumes that session->messenger exists, at the very least before the lock and under some condition even after it, and it also assumes that session exists throughout the whole function. If either session->messenger or session are deallocated by the time this function runs, there is more than just logger that needs to be fixed here.

[iphydf]

This is a generally better way, we should really not have the logger be a member of so many structs. Any of them could be initialised incorrectly or not at all. This pr is mainly to narrow down the possible error space of the linked issue.

[nurupo]

Sounds good.