[SYCL][UR][Bench] Run less variants of MemcpyExecute (#17456)

to speed up the benchmarks.

Performance difference between v1 and v2 is represented best by the
remaining variants.

Author: igchor

unified-runtime/scripts/benchmarks/benches/compute.py

@@ -74,14 +74,8 @@ def benchmarks(self) -> list[Benchmark]:
             ExecImmediateCopyQueue(self, 1, 1, "Device", "Host", 1024),
             VectorSum(self),
             MemcpyExecute(self, 400, 1, 102400, 10, 1, 1, 1),
-            MemcpyExecute(self, 100, 8, 102400, 10, 1, 1, 1),
-            MemcpyExecute(self, 400, 8, 1024, 1000, 1, 1, 1),
-            MemcpyExecute(self, 10, 16, 1024, 10000, 1, 1, 1),
+            MemcpyExecute(self, 400, 8, 1024, 100, 1, 1, 1),
             MemcpyExecute(self, 400, 1, 102400, 10, 0, 1, 1),
-            MemcpyExecute(self, 100, 8, 102400, 10, 0, 1, 1),
-            MemcpyExecute(self, 400, 8, 1024, 1000, 0, 1, 1),
-            MemcpyExecute(self, 10, 16, 1024, 10000, 0, 1, 1),
-            MemcpyExecute(self, 4096, 1, 1024, 10, 0, 1, 0),
             MemcpyExecute(self, 4096, 4, 1024, 10, 0, 1, 0),
             GraphApiSinKernelGraph(self, RUNTIMES.SYCL, 0, 5),
             GraphApiSinKernelGraph(self, RUNTIMES.SYCL, 1, 5),