bug#36609: 27.0.50; Possible race-condition in threading implementation

[dick . r . chiang]

[Standard disclaimer about not being a compiler semanticist]

> a single word...  change should ... be atomic.

I've never heard that.  I googled around; one helpful and only
mildly sententious monograph on the topic is

https://stefansf.de/post/qualifier-volatile-c/

Adding the volatile qualifier to `threads_holding_glib_lock` did not 
resolve my MRE on my particular architecture (whatever that may be).