MDEV-34178: Enable spinloop for index_lock

In an I/O bound concurrent INSERT test conducted by Mark Callaghan,
spin loops on dict_index_t::lock turn out to be beneficial.

This is a mixed bag; enabling the spin loops will improve throughput
and latency on some workloads and degrade in others.

Reviewed by: Debarun Banerjee
Tested by: Matthias Leich
Performance tested by: Axel Schwenke

storage/innobase/include/srw_lock.h

@@ -402,7 +402,7 @@ typedef srw_spin_lock_low srw_spin_lock;
 class ssux_lock
 {
   PSI_rwlock *pfs_psi;
-  ssux_lock_impl<false> lock;
+  ssux_lock_impl<true> lock;
 
   ATTRIBUTE_NOINLINE void psi_rd_lock(const char *file, unsigned line);
   ATTRIBUTE_NOINLINE void psi_wr_lock(const char *file, unsigned line);

storage/innobase/include/sux_lock.h

@@ -285,7 +285,7 @@ class sux_lock final
 typedef sux_lock<ssux_lock_impl<true>> block_lock;
 
 #ifndef UNIV_PFS_RWLOCK
-typedef sux_lock<ssux_lock_impl<false>> index_lock;
+typedef sux_lock<ssux_lock_impl<true>> index_lock;
 #else
 typedef sux_lock<ssux_lock> index_lock;