logger: Use thread_local instead of pthread_getspecific()

[rgacogne]

Short description

It might help with #6624, but feels cleaner to me anyhow.

Checklist

I have:

• read the CONTRIBUTING.md document
• compiled this code
• tested this code
• included documentation (including possible behaviour changes)
• documented the code
• added or modified regression test(s)
• added or modified unit test(s)

[tih]

So I tried this patch, and it did, indeed, fix the crash I described in #6624. It also points out another quirk: once the pdns_server runs without crashing, it's interesting how the very first thing it tries to log is the initialization of each module, even before it announces its general presence at all. This is unfortunate, as the modules try to log their initialization before the desired log level has been set, and so, the log announcements of their starting is lost.

[Habbie]

While losing the log messages is a pity, they mostly exist so we can spot mismatched pdns_server vs. dynamic backend module versions in bug report logs, so it's not a big problem.