Fix go/lem breakages on baseline configs

- Don't collect profiling info for now, as compiled code reference
  them directly.
- Only compile optimized after reaching baseline hotness threshold if
  tiered JIT is enabled.

Test: test.py, go/lem benchmarks.
Change-Id: I0d21d5f77825a710588ef5a7c11288a5b9757907
diff --git a/runtime/jit/profiling_info.h b/runtime/jit/profiling_info.h
index ada1036..14d76d2 100644
--- a/runtime/jit/profiling_info.h
+++ b/runtime/jit/profiling_info.h
@@ -130,6 +130,14 @@
     return MemberOffset(OFFSETOF_MEMBER(ProfilingInfo, baseline_hotness_count_));
   }
 
+  void SetBaselineHotnessCount(uint16_t count) {
+    baseline_hotness_count_ = count;
+  }
+
+  uint16_t GetBaselineHotnessCount() const {
+    return baseline_hotness_count_;
+  }
+
  private:
   ProfilingInfo(ArtMethod* method, const std::vector<uint32_t>& entries);