Do not run stackLevelSetter except for x86.

[sandreenko]

Stack level setter is required only for x86, on other platforms, it could be skipped or run only in debug.

However, there is an issue with disabling it:
Stack level setter has code that was supposed to be run only on x86 but is running on all platforms currently:

runtime/src/coreclr/src/jit/stacklevelsetter.cpp

Lines 328 to 340 in 5cded6d

	if (maxStackLevel >= sizeof(unsigned))
	{
	#ifdef DEBUG
	if (comp->verbose)
	{
	printf("Too many pushed arguments for an ESP based encoding, forcing an EBP frame\n");
	}
	#endif
	comp->codeGen->setFramePointerRequired(true);
	}
	}
	//------------------------------------------------------------------------

If you put it under TARGET_X86 as I did in #42197 you will see a big size regressions because you will have to encode bigger offset (from the rsp instead of saved rsp+offset):

-000013 lea      rbp, [rsp+C0H]
-000104 mov      dword ptr [V04 rbp+30H], r15d
-000108 mov      dword ptr [V06 rbp+40H], r12d
-00010C mov      dword ptr [V07 rbp+48H], r13d
-
+000102 mov      dword ptr [V04 rsp+F0H], r14d
+00010A mov      dword ptr [V06 rsp+100H], r15d
+000112 mov      dword ptr [V07 rsp+108H], r12d

and also see bunch of new bad LSRA decisions.

category:implementation
theme:throughput
skill-level:beginner
cost:small

[sandreenko]

Moving to future unless we have time after #7996 to do this.

[AndyAyersMS]

I am now co-opting this phase to also scan for unused throw helper blocks, so we will just be running it always.

[AndyAyersMS]

#95379 now uses this phase for trimming throw helpers. So, closing this.

TP hit from running this phase (non-x86) is around 0.2%