[Asan][Windows] Fix asan stack traces on Windows.

[akhuang]

Summary

While implementing inline stack traces on Windows (https://reviews.llvm.org/D88988)
I noticed that the stack traces in many asan tests included an extra frame.

Currently we get the PC and then do a stack unwind and use the PC to
find the beginning of the stack trace.
In the failing tests the first thing in the stack trace is inside an inline
call site that shouldn't be in the stack trace, so replace it with the PC.

This change prevents tests from failing if we have inline stack traces.

[vitalybuka]

can you please add a test?

[akhuang]

I'm not sure how to test, since the stack trace is only incorrect if we have inline stack frames, which we don't have at the moment (being implemeneted in https://reviews.llvm.org/D88988).

[rnk]

I think we have enough coverage. The commit message indicates that this will fix the test failures we observe when inline frames are added.

compiler-rt/lib/sanitizer_common/sanitizer_unwind_win.cpp

void BufferedStackTrace::UnwindSlow(uptr pc, u32 max_depth) {
   size = CaptureStackBackTrace(1, Min(max_depth, kStackTraceMax),
     (void **)&trace_buffer[0], 0);
   if (size == 0)
     return;
 
   // Skip the RTL frames by searching for the PC in the stacktrace.
   uptr pc_location = LocatePcInTrace(pc);
   PopStackFrames(pc_location);
+
+  // Replace the first frame with the PC because the frame in the
+  // stacktrace might be incorrect.
+  trace_buffer[0] = pc;
 }
 
 void BufferedStackTrace::UnwindSlow(uptr pc, void *context, u32 max_depth) {
   CHECK(context);
   CHECK_GE(max_depth, 2);
   CONTEXT ctx = *(CONTEXT *)context;
   STACKFRAME64 stack_frame;
   memset(&stack_frame, 0, sizeof(stack_frame));