_mjit_compile_pc_and_sp.erb: forget about memcpy [ci skip]

because using memcpy here actually made performance worse.
In Optcarrot,

* for loop
fps: 55.90428960597268
fps: 57.21690379342434
fps: 58.41751168279789
fps: 58.7317470613203
fps: 62.81371000202459
fps: 63.02541045417479
fps: 63.77324026366317
fps: 65.48067818347447
fps: 66.02324323831398
fps: 66.3816048847424
fps: 66.54526488660156

* memcpy
fps: 54.98389812126847
fps: 57.96524857391357
fps: 57.974918725327946
fps: 58.285987089087364
fps: 58.639003211333225
fps: 59.03244214601948
fps: 59.89964772199603
fps: 60.809490242786175
fps: 62.462070790624146
fps: 62.534599002918306
fps: 63.787312566163756

That's probably because the copied size is very small.

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64699 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

tool/ruby_vm/views/_mjit_compile_pc_and_sp.erb

@@ -21,7 +21,7 @@
                 rb_snum_t i, push_size;
                 push_size = -<%= insn.call_attribute('sp_inc') %> + <%= insn.rets.size %> - <%= insn.pops.size %>;
                 fprintf(f, "    reg_cfp->sp = (VALUE *)reg_cfp->bp + %ld + 1;\n", push_size); /* POPN(INSN_ATTR(popn)); */
-                for (i = 0; i < push_size; i++) { /* TODO: use memcpy? */
+                for (i = 0; i < push_size; i++) {
                     fprintf(f, "    *(reg_cfp->sp + %ld) = stack[%ld];\n", i - push_size, (rb_snum_t)b->stack_size - push_size + i);
                 }
             }