cpu/cortexm_common: remove breakpoint from hard_fault_handler

The hard fault handler would normally call core_panic() which in turn
calls ps() - this already helps debugging a lot of crashes caused by
a thread stack overflow.

With the breakpoint in place, the code will not advance to this, leaving
users unknown of this feature in the dark.

On top, the breakpoint is not very helpful if thread stacks are already
corrupted.

Let's just drop it for simplicity's sake - it's not there for any other
architectures.

cpu/cortexm_common/vectors_cortexm.c

@@ -449,7 +449,6 @@ __attribute__((used)) void hard_fault_handler(uint32_t* sp, uint32_t corrupted,
             : "r0", "r1", "r2", "r3", "r12"
             );
     }
-    __BKPT(1);
 
     core_panic(PANIC_HARD_FAULT, "HARD FAULT HANDLER");
 }