[DF] Distributed RunGraphs always submits concurrently

Making the decision on the scheduling mode based on `IsImpliciMTEnabled` would
mean forcing users to run `EnableImplicitMT` every time they want to submit
distributed graphs concurrently.

If the user has already chosen to run a distributed RDataFrame application,
RunGraphs should just submit all the computations concurrently to the
distributed scheduler. The "sequential" mode of operation can still be achieved
by e.g. calling `GetValue` on distributed proxies in a for loop.

bindings/experimental/distrdf/python/DistRDF/__init__.py

@@ -89,25 +89,18 @@ def RunGraphs(proxies: Iterable) -> int:
     """
     # Import here to avoid circular dependencies in main module
     from DistRDF.Proxy import execute_graph
-    import ROOT
-
     if not proxies:
         logger.warning("RunGraphs: Got an empty list of handles, now quitting.")
         return 0
 
     # Get proxies belonging to distinct computation graphs
     uniqueproxies = list({proxy.proxied_node.get_head(): proxy for proxy in proxies}.values())
 
-    if ROOT.IsImplicitMTEnabled():
-        # Submit all computation graphs concurrently from multiple Python threads.
-        # The submission is not computationally intensive
-        with concurrent.futures.ThreadPoolExecutor(max_workers=len(uniqueproxies)) as executor:
-            futures = [executor.submit(execute_graph, proxy.proxied_node) for proxy in uniqueproxies]
-            concurrent.futures.wait(futures)
-    else:
-        # Run the graphs sequentially
-        for p in uniqueproxies:
-            execute_graph(p.proxied_node)
+    # Submit all computation graphs concurrently from multiple Python threads.
+    # The submission is not computationally intensive
+    with concurrent.futures.ThreadPoolExecutor(max_workers=len(uniqueproxies)) as executor:
+        futures = [executor.submit(execute_graph, proxy.proxied_node) for proxy in uniqueproxies]
+        concurrent.futures.wait(futures)
 
     return len(uniqueproxies)