kill the slog_scope logger guard

[pjenvey]

When dropped, slog_scope's logger guard resets the global logger to 'NoGlobalLoggerSet' which panics on log messages. We introduced reset_logging in b2132f to avoid potential panics during tests but they can still occur (especially when compiled in debug mode which can enable more verbose logging in our support libs).

I think I'm seeing drop of this guard happen after our reset_logging is called. I'm not sure why, maybe because of our tie ins on the Python side, see the cffi ffi.gc calls.

I've also seen a random lazy_static! block initialize something in a thread that triggers panics. Long story short: they're difficult to avoid, so let's disable the guard for now and report this issue to slog_scope as bad behavior.