use atomic instead of mutex to allow multiple calls to disableLimitEnforcement

[MichaelRawson]

In #577 I failed to notice that FMB calls Timer::disableLimitEnforcement more than once for some reason. This causes FMB to hang on success (!). Oops.

Fix that by using an atomic flag rather than a mutex. This also seems more intuitive to me in retrospect.

[MichaelRawson]

After discussion with @quickbeam123 , probably std::mutex::try_lock is a better option. (or std::recursive_mutex)