Clarify using RMM with other Python libraries (rapidsai#1034)

Based on an internal question, clarified the README docs on what the APIs for enabling cuPy/Numba to use RMM simply inherit whatever the state of RMM is as opposed to modify the state (like enabling a memory pool).

Authors:
  - Jake Hemstad (https://github.com/jrhemstad)
  - Ashwin Srinath (https://github.com/shwina)

Approvers:
  - Bradley Dice (https://github.com/bdice)
  - Mark Harris (https://github.com/harrism)

URL: rapidsai#1034

Author: jrhemstad

README.md

@@ -684,6 +684,11 @@ allocations by setting the CuPy CUDA allocator to
 >>> cupy.cuda.set_allocator(rmm.rmm_cupy_allocator)
 ```
 
+
+**Note:** This only configures CuPy to use the current RMM resource for allocations. 
+It does not initialize nor change the current resource, e.g., enabling a memory pool. 
+See [here](#memoryresource-objects) for more information on changing the current memory resource.
+
 ### Using RMM with Numba
 
 You can configure Numba to use RMM for memory allocations using the
@@ -704,3 +709,7 @@ This can be done in two ways:
   >>> import rmm
   >>> cuda.set_memory_manager(rmm.RMMNumbaManager)
   ```
+
+**Note:** This only configures Numba to use the current RMM resource for allocations. 
+It does not initialize nor change the current resource, e.g., enabling a memory pool. 
+See [here](#memoryresource-objects) for more information on changing the current memory resource.