Make SIMD capability variables thread-local ...

... on platforms that support TLS, which should include all
currently-supported platforms
(https://libjpeg-turbo.org/Documentation/OfficialBinaries)

Addresses a concern raised in #87

Although it is still my opinion that the data race in init_simd() was
innocuous, we can now fix it for free thanks to
ae87a95861, so why not?

simd/powerpc/jsimd.c

@@ -2,7 +2,7 @@
  * jsimd_powerpc.c
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009-2011, 2014-2016, 2018, D. R. Commander.
+ * Copyright (C) 2009-2011, 2014-2016, 2018, 2022, D. R. Commander.
  * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
  *
  * Based on the x86 SIMD extension for IJG JPEG library,
@@ -41,7 +41,7 @@
 #include <sys/auxv.h>
 #endif
 
-static unsigned int simd_support = ~0;
+static THREAD_LOCAL unsigned int simd_support = ~0;
 
 #if !defined(__ALTIVEC__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
 
@@ -109,8 +109,6 @@ parse_proc_cpuinfo(int bufsize)
 
 /*
  * Check what SIMD accelerations are supported.
- *
- * FIXME: This code is racy under a multi-threaded environment.
  */
 LOCAL(void)
 init_simd(void)