JIT: Fix invalid memory barrier removal optimization (#91870)

The ARM32/ARM64 backends have an optimization where they optimize out
the latter of two subsequent memory barriers if no memory store/load has
been seen between them. This optimization should not be allowed to
remove memory barriers when a call has been seen.

Fix #91732

Co-authored-by: Jakob Botsch Nielsen <[email protected]>

Author: github-actions[bot]

src/coreclr/jit/emitarm.cpp

@@ -4850,6 +4850,7 @@ void emitter::emitIns_Call(EmitCallType          callType,
 
     dispIns(id);
     appendToCurIG(id);
+    emitLastMemBarrier = nullptr; // Cannot optimize away future memory barriers
 }
 
 /*****************************************************************************

src/coreclr/jit/emitarm64.cpp

@@ -8886,6 +8886,7 @@ void emitter::emitIns_Call(EmitCallType          callType,
 
     dispIns(id);
     appendToCurIG(id);
+    emitLastMemBarrier = nullptr; // Cannot optimize away future memory barriers
 }
 
 /*****************************************************************************