Remove call to cudaDeviceSetCacheConfig (#1536)

## Summary
No longer set CUDA device cache configuration to prefer L1 cache.  This does
not appear to affect any kernels in a negative way.  With more shared memory
available, reduction functions are faster in some cases.

## Additional background

## Checklist

The proposed changes:
- [ ] fix a bug or incorrect behavior in AMReX
- [ ] add new capabilities to AMReX
- [ ] changes answers in the test suite to more than roundoff level
- [ ] are likely to significantly affect the results of downstream AMReX users
- [ ] are described in the proposed changes to the AMReX documentation, if appropriate

Src/Base/AMReX_GpuDevice.cpp

@@ -415,9 +415,6 @@ Device::initialize_gpu ()
     AMREX_ALWAYS_ASSERT_WITH_MESSAGE(device_prop.major >= 4 || (device_prop.major == 3 && device_prop.minor >= 5),
                                      "Compute capability must be >= 3.5");
 
-    // Prefer L1 cache to shared memory (this has no effect on GPUs with a fixed L1 cache size).
-    AMREX_CUDA_SAFE_CALL(cudaDeviceSetCacheConfig(cudaFuncCachePreferL1));
-
     if (sizeof(Real) == 8) {
         AMREX_CUDA_SAFE_CALL(cudaDeviceSetSharedMemConfig(cudaSharedMemBankSizeEightByte));
     } else if (sizeof(Real) == 4) {