bstr: use memchr and memrchr for bstrchr and bstrrchr

May be faster to use optimized variants instead of simple loop.

meson.build

@@ -428,6 +428,8 @@ endif
 
 features += {'ppoll': cc.has_function('ppoll', args: '-D_GNU_SOURCE',
                                       prefix: '#include <poll.h>')}
+features += {'memrchr': cc.has_function('memrchr', args: '-D_GNU_SOURCE',
+                                        prefix: '#include <string.h>')}
 
 cd_devices = {
     'windows': 'D:',