dlopen() fixes (#929)

* Initial attempt to switch over to internally linked state.

* Add missing CMake update.

* hipLaunchKernelGGLImpl must be inline as well. Ensure internal linkage.

* Ensure global retrieval uses internally linked state.

* Hide HC in the implementation. Minimise ADL woes.

* Strange software exists, and must be catered to.

* Use a less spammy mechanism for ensuring internal linkage / non-export.

* Remove leftover internal detail.

include/hip/hcc_detail/macro_based_grid_launch.hpp

@@ -85,7 +85,7 @@ requires(Domain<K> ==
         hc::parallel_for_each(acc_v, d, k);
     } catch (std::exception& ex) {
         std::cerr << "Failed in " << __func__ << ", with exception: " << ex.what() << std::endl;
-        throw;
+        hip_throw(ex);
     }
 }
 
@@ -113,7 +113,7 @@ requires(Domain<K> == {Ts...}) inline void grid_launch_hip_impl_(New_grid_launch
                               group_mem_bytes, acc_v, std::move(k));
     } catch (std::exception& ex) {
         std::cerr << "Failed in " << __func__ << ", with exception: " << ex.what() << std::endl;
-        throw;
+        hip_throw(ex);
     }
 }